Non Technical Summary
Alfalfa, Medicago sativa L., is the most important forage crop in the USA. However, biomass yield improvement has been slow. We propose to investigate diploid alfalfa from three interrelated angles to better understand the genetic basis of biomass yield and identify possible genetic approaches to future yield improvement. First, we will develop highly accurate, haplotype resolved genome sequences for eight diploid genotypes from both alfalfa subspecies and resequence an additional eight genotypes to gain further information on structural variation among these individuals. Our sequences will include the first whole genome sequence of the yellow-flowered subspecies falcata. Second, we will investigate the biologically interesting phenomenon of segregation distortion toward excess heterozygosity. We will survey more than 20 segregating populations for excess heterozygosity, describing similarities and differences between the regions with distortion. Third, we will develop several large segregating populations that express heterozygote excess in some genomic regions and use them to genetically map quantitative trait loci (QTL) for yield and yield components. Finally, we will look for coincidence of QTL, loci with distorted segregation, and structural variants in the parental genomes to help explain variations in yield and propose mechanisms for how these loci control yield. These results will lead to genetic markers to assist selection for yield and potentially strategies to more effectively select for higher yield.

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
201	1640	1080	50%
201	1640	1081	50%

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

Subject Of Investigation
1640 - Alfalfa;

Field Of Science
1081 - Breeding; 1080 - Genetics;

Goals / Objectives
In this project, we aim to investigate four objectives:(1) To develop high quality, haplotype resolved reference genomes diploid M. sativa subsp. caerulea and subsp. falcata in order to evaluate structural variation,(2) To identify genomic loci exhibiting segregation distortion toward excess heterozygosity in diverse genetic backgrounds,(3) To map quantitative trait loci (QTL) for biomass yield and yield components, and(4) To compare the location of structural variants, loci exhibiting segregation distortion toward excess heterozygosity, and QTL for biomass yield and yield components.Our long term goal is to understand mechanistically the genetic drivers of biomass yield and to use that understanding to predict gene and/or haplotype combinations that will lead to yield improvement. This project is a first step toward that goal and will lead to testable hypotheses about loci controlling yield that can be incorporated into breeding programs.

Project Methods
Objective 1. We will sequence genomes of 16 genetically diverse diploid alfalfa genotypes using PacBio HiFi and Omni-C to generate phased genome assemblies of each genotype. These assemblies will be further improved by whole-genome paired-end 150bp short read sequences to correct any errors that may be present. We will extract high molecular weight DNA and develop SMRTbell libraries for High Fidelity (HiFi) sequencing based on PacBio protocols. The Omni-C libraries will be prepared using the Dovetail® Omni-C® Kit according to the manufacturer's protocol. Short-read PE150 libraries will be developed based on Illumina protocols. All sequencing will be done at the UC Davis DNA Technologies Core facility on a PacBio Sequel II system (HiFi), Illumina HiSeq (Omni-C), and a Novaseq S4 sequencer (Illumina PE150). We anticipate sequencing the entire alfalfa genome of ~800 Mbp to a depth of greater than 50×.We will generate de novo assemblies of these genotypes as follows. We will assemble HiFi reads using hifiasm (Cheng et al., 2021). Omni-C data will be assembled using AllHiC (Zhang et al., 2020). Omni-C and HiFi data will be combined using built-in functions of hifiasm for complete assemblies. Short reads (Illumina) will be aligned to assemblies using bowtie2 (Langmead and Salzberg, 2012). Alignments of haplotypes within and across genotypes will be performed with minimap2 (Li, 2018; Li, 2021; Kalikar et al., 2022). Single nucleotide and structural variation between haplotypes, both within individual genotypes and across genotypes, will be identified using the Genome Analysis Toolkit (GATK; McKenna et al., 2012) and SyRI (Goel et al., 2019).We will annotate genomes ab initio using Augustus (Hoff and Stanke, 2019). To improve annotation, we will use existing transcript data from our previously published experiments (Li et al, 2012) and the existing whole-genome sequencing projects. Further, we will also use isoform sequencing (Iso-seq) of the two genotypes for which gold standard reference genomes will be developed. We will collect leaves one week after plants had been clipped to a 3 cm stubble and from leaves and flower buds at about four weeks of regrowth when plants have begun to flower. Leaves will be collected on ice and frozen at -80C before extraction using Qiagen RNAeasy kits. cDNA libraries will be made by the UC Davis DNA Technologies Core facility and sequenced on the PacBio Sequel II system to obtain full length transcripts. Iso-seq data along with the previously published transcripts will be used to annotate assembled genomes using an established pipeline with IsoSeq3, Minimap2, StringTie, and Evidence modeler that has been used by the Monroe lab and others.Objective 2. To examine the prevalence of segregation distortion across diploid germplasm, we propose to evaluate distortion in S1 and F2 populations derived from the 16 parental genotypes sequenced above. These populations will include intra- and inter-subspecies hybrids, replicate F2 populations derived from different F1 genotypes from the same parents, reciprocal populations, and multiple populations for a given parent.For genotyping we will use the DArTag platform developed by Diversity Arrays Technology. Recently, the Breeding Insights program has developed two 3000 SNP arrays on this platform. The arrays were developed from SNP identified through resequencing a series of alfalfa genotypes from across the fall dormancy spectrum; thus, we anticipate these arrays will be useful for our purposes here. Plants will be grown in the greenhouse under typical conditions (~25C with a 16 hour photoperiod). Young, freshly expanded leaves free from disease and insect damage will be harvested, placed on ice, and freeze-dried. DNA extraction and sample preparations will follow the DArT instructions.For each S1/F2 population, we will initially screen 46 individuals and the parental genotype, with two populations per plate = 94 genotypes plus two slots for the DArTag controls. With this population size, our power to detect moderate deviations with a c2 test at a=0.05 is over 0.8. If the results from 46 individuals returns ambiguous results, we will genotype more individuals. However, at this phase of the project, our aim is simply to determine the universality of segregation distortion, particularly toward excess heterozygosity, so genotyping more populations is better. The DArT marker sequences will be aligned to the alfalfa genome sequences we have developed in Objective 1.The population sizes will be small to create genetic maps, but for the two larger populations, we will create genetic maps using QTL IciMapping (Meng et al., 2015), which has been used previously to identify SDR in rice (Liang et al., 2020), and JoinMap 5.0 (Van Ooijen, 2006; https://www.kyazma.nl/index.php/JoinMap/) to compare maps between populations. Segregation distortion regions will be defined as exceeding a LOD threshold of 3.0 in QTL IciMapping (Meng et al., 2015). Segregation distortion affects linkage analysis, so we will consider the use of the methods by Lorieux et al. (1995) to account for SD, as we did in Li et al. (2011a), if necessary.Objective 3. The individual genotypes of the populations will be heterozygous. To evaluate individual genotypes in the past, we have used vegetative clones via rooted stem cuttings for yield QTL mapping. Depending on the number of clones (or seeds) we have available, we will use either an a-lattice or an augmented design with replicated checks for field trials (Zystro et al., 2018; Burgueño et al., 2018). We will plant trials at the UC Davis Plant Sciences Farm in Davis, CA.We will measure biomass yield beginning in April or May the spring after establishment or when they begin to flower, which depends on weather and the level of dormancy these germplasm will show. We will harvest the plots at intervals when ~50% of plants have started to flower. Diploid germplasm tends to regrow more slowly than cultivated tetraploids based on my previous observations. Therefore, we expect to have 2-4 harvests per year.At each harvest period, we will score maturity based on the scale of Kalu and Fick (1981), measure the length of the tallest stem of each individual plant, harvest the center plant in the plot in one sample bag for additional measurements, and harvest all other plants in the plot into a second sample bag. Harvest will be done with hand sickles cutting at ~5cm from the soil surface. Both samples will be dried in a forced air drier at 60C, and the dry matter weighed. On the one plant subsample, we will separate leaves and stems, weigh each component, and compute a leaf:stem ratio. We will also count the number of primary stems on the plant.Phenotypic data analysis of linear mixed models will be conducted in AS-REML-R to generate best linear unbiased predictors (BLUP) to use in QTL mapping with JoinMap 5 for F2 populations or TASSEL (Bradbury et al., 2007) for the AIP. Broad sense heritabilities will be computed, and correlations among traits analyzed. A meta-analysis of QTL (Goffinet and Gerber, 2000) from our experiments and those of other alfalfa yield studies will be conducted using Meta-QTL (Veyrieras et al., 2007) to update the consensus mapping paper published by Ray et al. (2018)Objective 4. In this objective, which in some respects is accomplished as we go along with the others, we explicitly bring the three previous objectives together. Starting with the genomes of our parents, we will assess what differentiates them and determine if there are any significant features observed at the genome level - insertions, deletions, etc. - related to the regions where we have seen segregation distortion and to regions where QTL for biomass yield or yield components are located. Overlapping regions may be obvious by markers falling into QTL intervals and similar markers associated with multiple traits and/or segregation distortion.