Source: NORTH CAROLINA STATE UNIV submitted to NRP
IMPROVING GENOMIC AND PHENOMIC RESOURCES TO BUILD A FOUNDATIONAL BREEDING PIPELINE FOR DEVELOPMENT OF LOW-INPUT SEEDED ZOYSIAGRASSES
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
ACTIVE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1032284
Grant No.
2024-67013-42586
Cumulative Award Amt.
$270,300.00
Proposal No.
2023-11044
Multistate No.
(N/A)
Project Start Date
Jun 1, 2024
Project End Date
May 31, 2026
Grant Year
2024
Program Code
[A1141]- Plant Health and Production and Plant Products: Plant Breeding for Agricultural Production
Recipient Organization
NORTH CAROLINA STATE UNIV
(N/A)
RALEIGH,NC 27695
Performing Department
(N/A)
Non Technical Summary
While zoysiagrasses are recognized for their low input requirements and seeded cultivars are preferred by the industry for their cost-effectiveness and ease of distribution, a lack of understanding on the genetic components underlying seed yield, dormancy and germination have hindered breeding efforts. Several public and private programs have been working on the development of seeded cultivars; however, to date only two seeded cultivars are commercially available. In this project, we aim to develop a diversity panel for zoysiagrass consisting of approximately 250 genotypes to 1) understand ranges of genetic variation for seed traits, 2) increaseour understanding on the genetic components of seed yield and germination, and 3) initiate development of high-throughput phenotyping tools. We will perform genome sequencing and SNP calling on this diversity panel. Additionally, we will perform extensive phenotyping on seed traits, including germination rates, under controlled and field conditions. Genotypic and phenotypic data will be combined in a Genome Wide Association Study to identify genetic components underlying seed yield as well as germination efficiency. Additionally, we will investigate the use of imaging and machine learning techniques for counting seedheads and seeds. Data generated in this studywill provide a foundational understanding of the components of yield that could be used by breeders for making informed and more targeted selection decisions in breeding for seeded zoysiagrass cultivars.
Animal Health Component
60%
Research Effort Categories
Basic
30%
Applied
60%
Developmental
10%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
2012130108060%
2012130108140%
Knowledge Area
201 - Plant Genome, Genetics, and Genetic Mechanisms;

Subject Of Investigation
2130 - Turf;

Field Of Science
1081 - Breeding; 1080 - Genetics;
Goals / Objectives
In this project, we aim to develop a diversity panel for zoysiagrass consisting of 250 genotypes to 1) understand ranges of genetic variation for seed traits, 2) initiate development of high-throughput phenotyping tools, and 3) perform genome sequencing and SNP calling on this panel. Data will be combined for GWAS to increase our understanding on the genetic components of seed yield.Obj 1:Phenotypic evaluation of the zoysiagrass diversity panel for seed yield components under field conditions and germination rates under controlled conditionsObj 2:Preliminary assessment on the use of imaging and machine learning for high-throughput phenotyping of inflorescence and seedhead traits in zoysiagrassObj 3:Develop genomic resources for the zoysiagrass diversity panel
Project Methods
Obj 1:1) Field Evaluations: A germplasm panel of 250 diverse zoysiagrass genotypes including plant introductions, commercial cultivars and breeding lines from the zoysiagrass breeding programs at NCSU and TAMUS will be established in summer 2024 in Raleigh, NC and Dallas, TX. The trial will be arranged as partially replicated experimental designs (PREP) with 40% of the genotypes replicated. During the establishment year, the percent green cover will be evaluated using Unmanned Aerial Systems (UAS) to assess genotype persistence. In spring 2025, flowering time will be recorded and inflorescence abundance will be collected visually (1-9 scale, 1 = no inflorescences and 9 = 100% plot cover by inflorescences) and using digital imaging. Then, inflorescences will be harvested, dried, and imaged to measure inflorescence morphology including seed head length, width, area, and number of seeds per seed head. Seeds will be harvested and bulked by parent, and evaluated for final count, individual and average seed area, length, width, volume, surface area, and perimeter using the WinSEEDLE image analysis system (Regent Instruments, Inc.). 2) Germination Tests: Seed viability will be assessed by germination tests according to Qian et al. (2013). The number of replications and seeds/replication will be determined according to seed availability. All data will be analyzed using a mixed model approach, implemented in ASReml-R, in order to verify the significance of genetic effect and get genotype Best Linear Unbiased Estimators (BLUEs).Obj 2:Modern deep learning methods that utilize convolutional neural networks (CNNs) are quickly becoming a popular alternative to visual tasks that involve counting features or objects within digital images. To develop the model, images will be collected from a subset of genotypes under controlled lighting conditions within the greenhouse environment. Images will be collected weekly from the start of flowering to 100% seedhead maturity for all genotypes. Inflorescence and seedheads will be labeled using open-source AI-assisted labeling tools and used to train and validate the deep-learning model. Labels will differentiate seedheads from inflorescences and be used to track flowering time and seedhead development across genotypes. Seedhead morphology will be calculated from the polygon masks and validated against measurements from the WinSEEDLE image analysis system in Objective 1. To validate and compare models the datasets will be divided into training, validation, and testing, and performance assessed using appropriate evaluation metrics. Additionally, separate color photographs of field plots will be collected and used to train and test the ability to phenotype seedhead development and morphology at both the plot and field scale using a CNN approach.Obj 3:Genomic DNA will be extracted from genotypes in the diversity panel and sequenced via genotyping-by-sequencing (GBS) on Illumina NovaSeq 6000 SP 150bp PE lane at the NCSU Genomic Sciences Laboratory (GSL). Cleaned sequencing reads will be mapped to the Nagirizaki ZJN r1.1 reference genome using BWA. GATK and TASSEL pipelines will be used to discover SNP variation. Increased depth of coverage (>15) will be applied to avoid error rates during SNP calling in allotetraploid zoysiagrass, which has been successfully used in our previous GBS sequencing of zoysiagrass populations of similar size. Population structure and genetic diversity will be analyzed based on SNP data using PLINK. GWAS will be conducted using TASSEL 5 or other packages within R or Python, designed for mixed linear modeling (MLM), to identify associations for seed trait data collected in Obj 1.

Progress 06/01/24 to 05/31/25

Outputs
Target Audience:Audiences communicated with include breeders and seed procurement experts from the seed industry (f.e. Barenbrug, Scotts, MVP Genetics, among others), turfgrass scientists, county extension agents, master gardeners, lawn maintenance professionals, golf course superintendents, US Golf Association agronomists, and sports turf managers. Changes/Problems:Some of the genotypes in the panel have not flowered. However, this was expected as outlined in the proposal and is useful information for breeding programs (to avoid these materials as parents). What opportunities for training and professional development has the project provided?At NC State, two PhD and one MS students, two undergraduate research interns and one post-doc were trained in turfgrass breeding, seed physiology, field management, experimental design, molecular biology, and high throughput phenotyping through involvement with this project. At TAMUS, one post-doctoral researcher and one undergraduate student worker is being trained in turfgrass breeding, seed physiology, field management, data collection, analysis. How have the results been disseminated to communities of interest?A field day stop on this project was included in the 2024 NC State Turfgrass Field Day in Raleigh NC. Plots tours were given in Spring 2025 to US Golf Association agronomists, breeders from the seed industry, and seed procurement associates from seed companies. A poster summarizing these results was presented at the National Association of Plant Breeders (NAPB) meeting, where the PhD student working on this project engaged in discussions and received feedback from plant breeders and researchers. What do you plan to do during the next reporting period to accomplish the goals?Obj. 1: We plan to continue harvesting seedheads and phenotyping seedheads and seed morphology. Seed will then be evaluated for seed fill using Xrays followed by germination tests. All data from 2025 will be analyzed over winter. Another round of phenotypic evaluations will be conducted spring 2026. Obj. 2: For HTP, based on the promising results from the pilot greenhouse study, we plan to extend testing to the field using various sensors (RGB, multispectral, and hyperspectral cameras) and platforms (UAVs, Android phones, and mobile structures). These tools will enable fast, unbiased assessment of seed abundance, maturity, and color. The large volume of extracted data will be used to train deep learning models to predict flowering-related traits, improving prediction accuracy across a broader panel of genotypes and ultimately reducing the breeding cycle and accelerating selection. Obj. 3: Genomic libraries will be sequenced over the summer, followed by use of the TASSEL pipeline for SNP calling. Population structure and genetic diversity will be analyzed based on SNP data using PLINK. A preliminary GWAS will be conducted over winter using phenotypic data from year 1. ?

Impacts
What was accomplished under these goals? Obj 1: A germplasm panel that includes plant introductions, cultivars, and breeding lines from the Texas A&M and NC State University breeding programs was assembled in 2024. The 2024 GWAS was planted in Raleigh on 24 July 2024 and in Dallas on 26 Jul 2024 as single plugs on 1.2 m centers. A total of 269 genotypes were included at each site, though only 133 were replicated twice for a total of 402 plots. In May 2025, phenotypic evaluation of flowering and seedhead data was initiated at both locations. Data on flowering time, inflorescence color, seedhead density, and aerial images of all plots have been collected. Many plots have not flowered this spring. Harvest of seedheads for evaluation of seedhead and seed morphology will start next week. Obj. 2: To assess the relevance and feasibility of using machine learning models for flowering trait analysis, we designed and tested a high-throughput phenotyping pipeline. Results showed that seedhead abundance can be accurately annotated and predicted using a deep learning model. A poster summarizing these results was presented at the National Association of Plant Breeders (NAPB) meeting. Obj. 3: DNA was extracted from all entries in the panel and a library was constructed for whole genome sequencing.

Publications

  • Type: Conference Papers and Presentations Status: Accepted Year Published: 2025 Citation: Fratton, Stefano. AI meet zoysiagrass: Advancing seed production through high-throughput phenotyping. National Association of Plant Breeders, May 19-23, 2025, Kona, Hawaii. Poster #130. Available at https://napbannualmeeting.org/wp-content/uploads/2025/05/NAPB_2025_Booklet_Digital_VF2.pdf