LEVERAGING GENOMIC AND PHENOMIC SELECTION TO BREED BETTER COVER CROP AND FORAGE MIXTURES, FASTER

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: ACTIVE
• Funding Source: AFRI COMPETITIVE GRANT
• Reporting Frequency: Annual
• Accession No.: 1030313
• Grant No.: 2023-67014-39449
• Cumulative Award Amt.: $300,000.00
• Proposal No.: 2022-10260
• Project Start Date: Sep 1, 2023
• Project End Date: Aug 31, 2025
• Grant Year: 2023
• Program Code: [A1141]- Plant Health and Production and Plant Products: Plant Breeding for Agricultural Production

Recipient Organization
AUBURN UNIVERSITY
108 M. WHITE SMITH HALL
AUBURN,AL 36849

Performing Department
(N/A)

Non Technical Summary
Intercropping is a diversification practice in which two or more crop species or varieties are grown in the same place and/or time. When optimized, intercrops provide higher, more stable overall yield and environmental benefits like erosion control and weed suppression. Intercropping research has mostly been concerned with agronomic management and has used existing crop cultivars, developed in monocultures. Many studies demonstrate that intercrop conditions represent a different environment and that variety choice matters; this strongly implies that systematic breeding will help achieve their full potential. However, breeding for intercrops is a massive combinatorial problem, which has limited the development of intercrop-adapted cultivars. In a phenotypes-only selection paradigm, it is feasible to integrate evaluation for performance in mixtures only into the later stages of a breeding pipeline.Intercrop selection needs to be applied recurrently and in early stages when there are 100's-1000's of each species to test. Genomic selection (GS) and aerial imaging with UAVs are often and successfully used to accelerate monoculture breeding when there are too many combinations to test. By genotyping all entries and phenotyping representative subsets, a reference dataset is built and used to predict the performance of breeding lines even if they have not been phenotypically evaluated.Our focal intercrop is a dual-purpose forage and cover crop mixture of crimson clover (Trifolium incarnatum) and oats (Avena sativa). Over four location years, we will generate foundational data to demonstrate a novel use of genomic and phenomic prediction, selection for intercrop-ability and overall performance in early-stage breeding trials.

Animal Health Component

20%

Research Effort Categories

Basic

60%

Applied

20%

Developmental

20%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
201	1699	1081	100%

Knowledge Area
201 - Plant Genome, Genetics, and Genetic Mechanisms;

Subject Of Investigation
1699 - Pasture and forage crops, general/other;

Field Of Science
1081 - Breeding;

Keywords

cover crops

forages

genomic selection

intercropping

phenomics

plant breeding

Goals / Objectives
Our main objective is to establish foundational populations and data to demonstrate how genomic selection and phenomics can overcome an otherwise intractable combinatorial problem, that of breeding two populations for intercrop-ability and performance.Obj. 1: Determine the potential to improve mixture performance by breeding in either one or both species by assessing the general and specific mixing ability of early-stage breeding lines using genomic prediction.Obj. 2: Develop scale-able, UAV-based phenomic systems that can measure production over time and distinguish changes in the relative biomass compositions of mixtures.Obj. 3: Assess the accuracy of genomic and phenomic predictions of the performance of mixtures and estimate the acceleration in intercrop improvement expected from combining one or both approaches.

Project Methods
PIs Wolfe and Rios will jointly run crimson clover and oat trials. Clovers are obligate-outcrossing cool-season annuals. We will test 200 half-sib clover families against 30 oats per year at two locations (AL+FL). The pilot trials will emphasize crimson clover over oat but will generate estimates of the mixing ability variance for oats to guide future investment in each species (Obj. 1). This year, pre-funding, we are doing seed increases and preliminary evaluations of crimson clover and oat populations so by project start (Fall '23) we will be ready to evaluate populations in diverse mixture plots. Figure 2 illustrates the GS scheme to-be-tested. Two trial types are run in tandem at each location:Stage 1: Spaced plantings for genotyping and selection.Purpose: Grow out and select the best individual plants as progenitors of the next generation.Details: Plant in isolation from other T. incarnatum. 20 half-sib families/year, 6 plants/each x 4 reps = 480 plants, 1.5' spacing. Traits-to-score: emergence, vigor, flowering (time, height), dry biomass, seed yield. Selection: Cull at flowering. Keep 300 most vigorous plants. After harvest, select the top 200 for next year's legume-grass mixtures (Stage 2), based on biomass, seed yield, and maturity group (early, late). Select the top 20 for next year's spaced-plant evaluation. Genotyping: Collect and freeze tissue from all plants but genotype only the 200 selected plants.Stage 2: Grass-legume mixture plots.Purpose: Evaluate breeding populations in mixtures to generate training data for GP.Details: Plant 400 grass-legume mixture plots per year per location with a plot drill, 5' wide x 10' long. 200 clover vs. 30 oat entries, 2 different oats per clover per location-year, more combinations, and replication as seed availability allows. Recommended seeding rate: 10 and 30 lb/acre for clover and oats, respectively. Traits-to-score: See phenotyping/phenomics below. Selection: No selection. Data for genomic prediction. Genotyping: Clover: family-pool genotyping (see below). Oats: are inbreds, and can be genotyped just once.GS for Mixing Ability: In Year 2 we will apply GS for GMA-SMA. We will break the Stage 1 trial into two parts. Stage 1a: Establish Spaced Plants as usual. Genotype all plants. Train prediction model using previous year(s) mixture plot data. GMA predictions are used to select the best 200 plants for next year's mixture plots. Stage 2b: Pollination Nursery of Predicted Best Plants. Dig up the top 20 plants. Replant in a pollination cage for controlled crossing to increase selection intensity and genetic gain. Models are detailed in the "Data analysis" section.Genotyping: As an obligate outcrossing species, each clover-oat mixture plot will contain a half-sib. clover family. We will use family-pool genotyping - combining tissue from multiple plants per family to estimate allele frequencies. PI Rios et al. have shown this approach works for GS in Alfalfa and other forages [80-85]. We will train models to predict the mixing ability of individually genotyped clover plants in spaced plant trials (Stage 1 - Year 2) based on crimson-oat mixture plot phenos (Stage 2 - Year 1+2) and their corresponding family-pool genos.Phenotyping / PhenomicsObjective 2 is to develop high-throughput, non-destructive phenotyping of mixtures. UAVs with multi-spectral sensors have been effectively used in many species, including in forages by PI Rios [86] to non-destructively predict biomass, dry matter, height, ground cover and N-content [87-90]. PI Bao (Auburn) and PI Rios (UF) will image our mixture plots with a multi-spectral camera-equipped drone, at least monthly.Ground-based non-destructive measures: At each flyover, we will take measurements to pair with UAV data. Traits: phenology, height, % clover-vs.-oat composition, canopy light interception (leaf area index meter), and N-content (chlorophyll meter).Destructive measures: At least once during the season, we will harvest, botanically separate and record fresh and dry biomass by species, from a 1-m2 section of each plot. UAV images before and after will quantify regrowth capacity.Nitrogen Fixation, Water Use Efficiency, and Forage Quality: Biological fixation of atmospheric N (BNF) by legumes is a critical reason to include them in mixtures. Biomass analysis of 15N isotope concentration quantifies the relative amount of N derived from BNF [91] and has been used to identify differences in fixation between crimson cloversand by PI Sanz-Saez in soybeans [94]. Forage quality traits including dry matter, C:N ratio, protein, and fiber content will be measured by near-infrared spectroscopy (NIR) [95]. At 50% flowering, we will sample biomass from each mixture plot. One sample will be sent to UF for NIR-based forage quality determination. Another tissue sample will be separated by species and sent for isotopic analysis. 15N in clover samples will be compared to paired, non-fixing oat samples. We will also have carbon isotopes (13C:12C) quantified as a measure of water use efficiency as it has been previously done by Co-PI Sanz-Saez in soybean, common bean, and peanut.Final above and belowground measurements: At full maturity, we will harvest, separate by species, dry, and weigh the biomass of each plot. PI Sanz-Saez will dig up and wash clover+oat root crowns (4 plants/plot) and use his RhizoVision imaging system [99] to examine root density and architecture.Data analysis and interpretationGS is not a new technology, but we are putting it to new use. GP for intercropping has been described in the literature [8,66]. We can adapt approaches for similar situations, including predicting GxE [71], hybrid breeding (GCA-SCA) [77] and direct-indirect genetic effects in animalsand use existing mixed-model software.Interpretation and Relationship to Project Objectives:Objective 1 is to determine whether mixture performance can be improved by breeding in either one or both of the species. We will estimate the GMA and SMA fractions of genetic variances to determine how to invest our future breeding efforts. The relative variability for performance in each mixture of each species will indicate the improvement of intercrop performance expected from selection in that species. The relative variability for performance of specific mixtures will indicate the advantage gained by joint selection in both species.Objective 2 is to develop UAV-based phenomics for assessing non-destructively and over time, traits useful for selecting better species mixtures. We will estimate the genetic correlation between ground-based measures, especially the destructive methods, using e.g. multi-trait mixed modeling. We will identify critical timepoints and vegetation indices where UAV-based imaging can best predict traits like final biomass. Long-term, we envision using these data to develop AI-based delineation of species biomass composition and other traits.Objective 3 is to assess the accuracy of genomic and phenomic predictions of mixture performance and to estimate the acceleration in intercrop-improvement expected from using either or both approaches combined. We will conduct cross-validation analysis to estimate prediction accuracies for GMA in each species, the SMA and overall mixture performance. We will compare the accuracy of UAV-based phenotypes, genomic prediction and their combination for predicting biomass and the other traits. Lastly, we will estimate the rate of intercrop improvement expected using the breeder's equation and stochastic simulations.

Progress 09/01/23 to 08/31/24

Outputs
Target Audience:Our goals support public sector breeding, lead to the enhancement of germplasm, and apply quantitative genomics and phenomics to understand the genetics of performance in mixtures. Public and private sector plant breeders and geneticists are the primary audiences for the scientific goals of our project. In addition, soil and crop scientists in the areas of cover crops and forages are also likely audiences, as this is our focal system for breeding mixtures. In the long run, we hope to develop improved cultivars and mixtures, and therefore, seed companies and farmers are also stakeholders/audiences for this project. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Two graduate students are supported under this project: Gideon Oyebode (AU) and Sandip Aryal (UF). For Gideon, the project will be the basis of 2 thesis chapters/publications. Gideon has had the opportunity to present his work at two conferences. Several (up to 6) different undergraduate students have worked on this project as technical support (processing samples, etc.). How have the results been disseminated to communities of interest?So far, the work have only been disseminated as posters presented atconferences (NAPB, CROPS2024). What do you plan to do during the next reporting period to accomplish the goals?In Year 2 (of 2) we will complete our bioinformatics pipeline and spend most of the year assessing genomic variances (Obj 1) and prediction accuracy (Obj 3). We will complete the processing of samples and obtain and begin analyzing the NIR, Isotope, and drone data for Obj. 2. We've just planted the second year of field trials and will phenotype and genotype these as planned to complete the datasets we set out to generate.

Impacts
What was accomplished under these goals? This project seeks to overcome the current limitations on breeding for mixtures by using genomic and prediction of GMA (Obj. 1+3) and phenomic tools (Obj. 2). Genomic prediction of GMA prospectively can be done without having to observe all genotypes in actual field combinations. Instead, we proposed to use genotyping of all members of both species and to field test combinations "sparsely," meaning that all individuals of both species are tested with some companions from the other species, but a more reasonable subset than testing all-against-all. Genomic information will link plots together based on relatedness. This is modeled after how hybrid crop breeding and multi-location GxE trials are done. The proposed focal cover crop mixture is Crimson Clover (Trifolium incarnatum, CC for short) and Oat (Avena sativa), which has additional uses as forage. Pre-project Year 0 (2022-2023): The field trial proposed for this project derived seed from the "Space Plant Nursery" trials at AU and UF during 2022-2023. In 2022, we at AU started crimson clover breeding from scratch; UF's program has been under PD Rios' for >5 years. At AU acquired the entire NPGS germplasm collection (42 lines) and grew it alongside an AU population that had been stored since 2016. In September 2022, 350 crimson clover plants, up to 10 per accession, were germinated in a greenhouse and were transplanted to the field about six weeks later. Plants were allowed to open-pollinate to generate variability. Seeds were harvested by hand at maturity in the first two weeks of May. Initial seed yield is our only selection criterion at this point. The ability to produce enough seed will be the first criterion for our ability to breed (and produce seed) in our local climate. Of the original 350, 105 plants provided seed. Of the original 42 NPGS accessions, 34 yielded seed. The AU plants outperformed all the rest with an avg. 3681 seed/plant (76% more than Dixie, which is the most common commercial variety). Dixie plants made an avg. of ~2045 seed/plant. The NPGS collection had a similar avg. (2071 seed/plant), but being a diverse collection, accessions ranged in performance up to almost 6.5K seed/plant. Seed production in 2022-2023 revealed that we would only have a few grams of seed per plant to work with. Therefore, we could only plant single-row plots rather than the multi-row drilled plots we had hoped for. Perhaps compensating for this, we were able to use more entries and have more total plots than originally planned. Project Year 1 (2023-2024): In 2023-2024, we planted 2 trials. Trial 1: Crimson Seed Increase Nursery. Also called a spaced plant trial (SPT). This year, we planted two nurseries at EVS-PBU, starting with 1722 seedlings at PSRC. These represented 188 maternally related half-sib lines from 2022-2023, including 82 new entries shared by Rios at UF. They were allowed to open pollinate to maximize diversity and trait combinations before future selections. After the 2023-2024 season, there are 794 single-plant lineages of seed, in contrast to 105 in the previous year! We selected 90 early and 90 late maturing lines for two nurseries in AU for 2024-2025. Trial 2: Crimson-Oat Early-generation Yield-in-Mixtures Trial. We designed a two-location trial with 600 plots at each location, 1200 total. Based on seed yields from 2022-2023, we determined that the best available plot configuration was a single 6ft long row. Interow space was 36" to allow between plot weed control. There were 179 Crimson Clover entries. 26 Oat lines were sourced from SunGrains breeders Steve Harrison (LSU) and Ali Babar (UF). The experiment is classified as a partially-replicated design within and across locations (aka p-rep, p-loc). CC-Oat combinations were chosen such that: Almost all clovers and all oats are observed in both locations All clovers observed with one or more oat All oats observed with one or more crimson 75% of plots are unique combinations of clover-oat (unreplicated combos) 15% of plots have combos replicated across-location 10% of plots have combos replicated within-and-across This led to 961 unique combos or ~21% of the 4654 possible combinations observed in the field. The trial was planted at EVS-PBU on November 8th, 2023. Phenotyping: The most important trait collected is the final biomass, collected at target cover crop termination time, in this case mid-April. Biomass was separated by species, dried, and weighed. Samples are currently being ground such that (1) crimson clover tissue can be tested for the amount of biological nitrogen fixation, and (2) mixed samples can be tested by NIR for "forage nutritional quality" parameters. We flew our Mavic 3M multispectral drone approx. every 3rd week and will use the data to extract plot biomass/volume estimates and pair them with ground-truthed measurements. We also manually recorded per species canopy height and maturity status several times during the season. Crimson Clover Genotyping: Crimson clovers are outbred and open-pollinated. We collected a leaf tissue sample from up to 20 plants per accession per location. These tissues were combined into a single pooled sample per accession per location. Samples were sent to HudsonAlpha for low-pass sequencing and imputation using their Khufu program on June 3rd, 2024. In August 2024, initial genotyping data were released to us for analysis. 439 samples were successfully sequenced (~50 failures). 168260 SNP were called and imputed based on the Khufu pipeline. Samples were aligned to the Red Clover reference genome (we have a draft Crimson Clover genome in progress). Challenge: Our samples are of populations, not individuals. We are sequencing to capture allele frequencies, not genotypes. The Khufu data product is not designed for that and, therefore, is likely not suitable for our use. We made the first attempt at analysis using the Khufu dataset, as is. We tested whether accessions that were genotyped twice (from AU and UF plots) appear to be nearly identical or similar. Unfortunately, this was not the case. If the results had been similar, we would have used the Khufu pipeline as is. Instead, we will develop our own bioinformatic pipeline to make use of the sequence data. We have initiated these efforts and are communicating with the Khufu team about possibilities for help. Significant genetic variation and prediction ability for Crimson Clover biomass yield in mixtures! For brevity, we don't include the model details or results here completely. We fit a linear mixed model using asreml-R 4.1. The model included a main effect for CC accession (corresponding to CC GMA), a main effect for Oat (Oat GMA) and an interaction effect (CC x Oat, aka specific mixing ability). The genetic relatedness of clovers-to-clovers and of oats-to-oats is factored into this model computed by the SNP marker data. Our results are modest but promising! The CC effect contributed significantly to the model, explaining about 12% of total variance. 25 reps of 5-fold cross-validation gave us an estimated prediction accuracy of 0.06; significantly different from zero, but certainly not a powerful prediction machine yet. We have lot's more data to evaluate and serious improvements in the SNP data (as outlined above) yet to do. We also used a separate analysis of CC and Oat yields to break down CC mixing ability into a Producer (clover effect on clover yield) and an Associate-effect (clover effect on oat yield). The clover Producer-Associate effects have a genetic correlation of -0.18: more vigorous clovers yield more but also cause reduced oat yields. There is nevertheless a large margin to select clovers with above average yield with lesser impact on oats yield We planted another 2 locations x 600 plots in November 2024, this time with 12ft long plots. We included 200 entries, including 29 "advanced lines" from the Cover Crop Breeding Network.

Publications

• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Oyebode, G., Aryal, Sandip., Bao, Y., Sanz-Saez, A., Rios, E., Wolfe, M.D. Genomic selection in intercrops: A pilot study of grazeable clover-oat cover crop mixtures. National Association of Plant Breeding. July 2024.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Oyebode, G., Aryal, Sandip., Bao, Y., Sanz-Saez, A., Rios, E., Wolfe, M.D. Genomic selection in intercrops: A pilot study of grazeable clover-oat cover crop mixtures. CROPS2024. June 2024.