Performing Department
Plant Pathology
Non Technical Summary
Abiotic stresses severely limit agricultural productivity. Often approaches to understand stress response miss the connectedness of complex biological functions. The primary goal of this proposal is to further our understanding of abiotic stress response dynamics by integrating phenotyping, ionomics, and transcriptomics data together in maize. We hypothesize a fuller understanding of stress responses will be achieved by researching across different abiotic stresses using large multi- omic datasets (phenomic, ionomic, transcriptomic). This project has three main objectives: 1) Determine the heritability of hyperspectral signatures and morphological traits under related stress environments to test if these phenotypic traits are indeed heritable as would be necessary for breeding purposes; 2) Integrate high-information phenotyping, ionomics, and global gene expression data to understand physiological, morphological, elemental, and transcriptomic response of stress across seedling development holistically; and 3) Develop and test predictive models for stress response from kernel to seedling. This objective will model kernel trait information to predict stress responses in seedlings. Researching plant stress response dynamics across scales will allowing modeling of stress response behavior and towards predictions at different developmental stages, which will help to increase breeding gains much like selections based on genotypes. Through this collective project, we will be linking multiple large datasets together to improve crop productivity under changing growing conditions. This project will produce several outputs including, well-annotated public datasets and phenomics tools beneficial to the maize, phenomics, and agricultural communities. It will further our understanding of plant response dynamics across scales and model stress response behavior, which will enable community work toward predictions of stress response at different developmental stages to increase breeding gains, much like selections based on genotype do. This project is well aligned with priorities of program area A1152 to "improve productivity...of agriculturally-important plants using molecular...approaches" and support research in "mechanisms of plant responses to abiotic stresses" and "nutrient uptake/utilization".
Animal Health Component
0%
Research Effort Categories
Basic
50%
Applied
50%
Developmental
0%
Goals / Objectives
The ability of crop plants to respond to and overcome environmental stress with minimal effects on yield and quality is critical to ensure future food security. A priority for program area A1152 is to improve productivity or other performance factors using molecular, biochemical, whole-plant, agronomic, or eco-physiological approaches and to support research in mechanisms of plant response to abiotic stresses. The primary goal of this research proposal is to extend our understanding of the dynamics of abiotic stress response by integrating multiple big data approaches including, high-information phenotyping, ionomics, and transcriptome analysis in maize. The central hypothesis of this project is that an integrated multi-scale approach to examining stress response mechanisms will provide a more holistic view of how stress response mechanisms work, in comparison analysis of individual stresses. Understanding plant stress response dynamics across scales will allow modeling of stress response behavior. Developing models of stress response behavior will help set the stage for community work towards predictions of stress response behavior at different developmental stages, which will increase breeding gains much like selections based on genotypes. To achieve this overall goal the specific objectives of this proposal are as follows:Objective 1: Determine the heritability of hyperspectral signatures and morphological traits under related stress environments. High-information phenotyping methods, such as hyperspectral imaging and temporal visible imaging, are becoming more broadly used in plant science. For these technologies to be useful in breeding programs the information extracted from these data types needs to be heritable. This objective will determine the heritability of information extracted from hyperspectral and visible images for a maize diversity panel under stress environments that are likely to co-occur.Objective 2: Integrate high-information phenotyping, ionomics, and global gene expression data. The goal of this objective is to understand the physiological, morphological, elemental, and transcriptomic response of stress across seedling development. This aim will involve trait collection, extraction, and the development of integrated network analyses to combine connected, but distinct datasets.Objective 3: Testing the predictive power of an integrated multi-omics approach for stress response from kernel to seedling. In this objective we will model and predict maize seedling stress response by integrating multiple phenomic and ionomic assays of maize kernels. Through this objective we will use extensive kernel trait information in models to predict stress responses uncovered in Objective 2 at the seedling stages. If successful, this objective could result in quick, cheap, and novel methods to predict seedling development based on kernel traits.
Project Methods
Objective 1: Selection of germplasm for this project: Wisconsin Diversity Panel. For this project objective to be successful we need a panel of maize lines with variation in response to multiple stress conditions. The germplasm we will use in this proposal is from the maize Wisconsin Diversity Panel. This panel of lines was put together at the University of Wisconsin-Madison and consists of a set of 627 lines that were selected based on flowering time, trialing ability, and maximum diversity based on pedigree information. This panel has been previously phenotyped in field trials and genotyped. In addition, a subset of the panel was used to determine genome content and gene expression variation. This population has been used extensively in the maize community as a resource to characterize natural variation for a large number of traits, numerous association mapping studies, and the variation present in the population has always been sufficient for any trait that has been collected to date.Application and collection of stress response data for Wisconsin Diversity Panel. The selected lines from the Wisconsin Diversity Panel will be phenotyped under four controlled abiotic stress conditions (reduced water capacity, high temperature, low nitrogen, low phosphorus) and standard control conditions. For all environments, exposure to the control and sub-optimum growth conditions will start at germination and continue for two-weeks of growth. Plants will be imaged daily with visible and hyperspectral images collected on alternate days. Seedling stage data collected through this proposal will be compared to available field data that includes measurements such as height, stalk diameter, 300 kernel weight, and leaf number to determine if adult traits are at all indicative of seedling stress response.Assessment of stress response by visible and hyperspectral imaging. Analysis of visible images collected in the 5 environments each will be done with PlantCV, which has been used to analyze plants of diverse architecture types, including maize, for an array of traits including, color, biomass, growth rate, water-use-efficiency, and height. For visible image data, analysis will focus on quantification of stress as well as growth traits. Hyperspectral images will be processed by previously developed in-house pipelines. Once plant-level data is extracted from hyperspectral images, well-characterized hyperspectral wavebands and indices such as NDVI, normalized pigment chlorophyll index, greenness index, photochemical reflectance index, and nitrogen and phosphorus content indices will be used to analyze plant responses to the 5 environmental conditions used in this project.Assessment of heritability. Following the extraction of morphological and spectral traits from hyperspectral and visible images, heritability will be calculated. Calculating heritability is important to determine if extracted traits can be used for selection breeding.Objective 2: Integrate high-information phenotyping, global gene expression, and bionomics data. Transcriptomics/ionomics sampling. At the end of the two weeks of imaging, tissue will be collected for transcriptomics and ionomics. Special care will be taken to ensure that sample collection is restricted to the same time after dawn (Zeitgeber time) for all experiments. Samples will also be collected from leaves of the same developmental age.Integration of phenomics, transcriptomics, and ionomics data: To integrate datasets together we will use networks that represent different omic layers. A network in this sense is a graph that consists of nodes and edges. Nodes in our case will be represented by gene expression, elemental profiles, and phenotypic traits (both morphological and hyperspectral). The edges in the networks then represent either a physical or functional relationship between nodes. Altogether, integration of phenomics, ionomics, and transcriptomics data will identify genes associated with abiotic stress and provide more information about gene function. This network approach utilizes pattern matching of frequently co-occurring items and is a type of association-based guidance method. There are other association-rules-mining algorithms that have been developed for integrating multiple omic datasets together that use similar but distinct approaches. These methods are constantly evolving, and we plan on testing and developing several analysis methods for the data collected. In general, most of the techniques we will use will take one of two approaches. A top-down data reduction approach would use the transcriptomics data as a basis to predict ionomic, spectral, and morphological changes. The benefit to this approach is that transcriptomic data coverage is greater so changes in regulation, transport, and other functions may be more easily captured. The alternative approach, bottom-up, would use the ionomics data as the starting point to guide the other 'up-stream' omics analysis.This starting points of this approach are quick and cheaply generated, but the coverage (elements quantified) is limited which reduces the biological search space There are several open-source programs that can be utilized as starting points for this analysis including, OmicsPLS, mixOmic, OmicKriging. We will use these tools to uncover associations in phenomic, ionomic, and transcriptomic datasets to better understand the holistic mechanisms of plant stress response.Objective 3: Testing the predictive power of an integrated multi-omics approach for stress response from kernel to seedling. Multi-omics assays will be applied to test the power of kernel phenotyping to predict and model our cataloged seedling stress response. Models to predict seedling stress response based on kernel phenotypes will be developed.Kernel ionomics and phenotyping of 50 Wisconsin Diversity Panel lines. We will conduct ionomics on kernels of the 50 selected lines from the Wisconsin Diversity Panel used in Objective 1. To complement elemental profiling of kernels, image-based and hyperspectral kernel phenotyping will also be conducted. For image-based data we will utilize an established automated pipeline designed specifically for maize kernels that returns several kernel traits, such as length, depth, width, contour, and area. We will also collect hyperspectral images from kernels of the 50 selected lines. Ideally, maize kernel phenotypes (morphological, spectral, and ionomic) will be used to predict seedling responses to abiotic stress (morphological, spectral, ionomic). To this end, models will be iteratively generated and tested using machine learning algorithms. Methods to be tested will range from simple linear regression models to more complicated methods such as support vector machines that have been previously used to generate predictive models for abiotic stress. Phenomics data for a randomly selected subset of the 50 phenotyped lines (kernel and seedling data) will be used as a training set, and the remaining lines will be used to test the resulting models. Testing and training will be done iteratively.