Performing Department
(N/A)
Non Technical Summary
Blackberry growers and stakeholders from every major production region in the US cited the lack of consistently flavorful cultivars as the most important constraint limiting the growth of the industry. Selection for improved flavor is a slow and inefficient process in blackberry breeding programs. Novel selection methods for improving flavor have been developed in other fruit crops, which are more accurate and objective than breeders' ratings and more affordable and scalable than traditional consumer sensory panels. These methods include the use of molecular markers for specific volatiles and acids and the implementation of genomic and metabolomic selection to predict consumer sensory ratings. We propose to leverage new genomic tools developed in blackberry and novel approaches for improving flavor developed in other fruit crops to develop improved selection methods for flavor in blackberry breeding programs. Specifically, we propose to combine data from consumer sensory panels and metabolomic data of volatile and nonvolatile factors to determine the most important breeding targets that impact consumer liking of blackberries. Then we will conduct GWAS to identify genomic regions controlling those volatiles, organic acids and sugars chosen as important breeding targets. Last, metabolomic and genomic selection models will be developed and compared to predict the consumer sensory ratings of new breeding selections. New genetic markers and optimized breeding strategies developed in this project will be deployed in applied blackberry breeding programs to cull seedlings from breeding populations, choose optimal parent combinations for crossing, and select candidates for advanced trials
Animal Health Component
0%
Research Effort Categories
Basic
30%
Applied
70%
Developmental
0%
Goals / Objectives
Our long-term goal is to develop flavorful cultivars that drive increased consumer demand for blackberries. The primary objective of this proposal is to optimize methods (marker assisted selection, genomic selection, metabolomic selection) for predicting consumer preference in the UA fruit breeding program in order to reduce breeding cycle time and expedite the process of combining flavor with other desirable traits. We hypothesize that sweetness, acidity, and a few key volatiles all play important roles in consumer preference and that selections with optimal flavor can be fast-tracked for parent use and advanced testing using metabolomic signatures and genomic data. The specific objectives of our proposal are:Objective 1: Assess the impact of volatile and nonvolatile factors and the influence of genotype and environment on flavor and consumer preference in blackberry. We will measure basic composition, acids, sugars, and volatiles and conduct consumer sensory evaluations of a large panel of fresh-market blackberry genotypes to determine the most important drivers of consumer preference for blackberry flavor and measure the percent variance of consumer sensory and biochemical attributes explained by genetic and environmental components and their interaction.Objective 2: Conduct genome-wide association mapping of organic acids, sugars, and volatile compounds in blackberry selections from the UA breeding program. We will phenotype and genotype a panel of 245 fresh-market blackberry selections and conduct GWAS to find marker-trait associations for key breeding targets associated with consumer preference in blackberry that were identified in objective 1.Objective 3: Compare genomic and metabolomic selection models for predicting flavor and consumer preference in blackberry. We will use the phenotypic and genetic data generated in Objectives 1 and 2 to develop and compare genomic, metabolomic, and combined selection models for flavor and consumer preference.
Project Methods
Objective 1: Assess the impact of volatile and nonvolatile factors and the influence of genotype and environment on flavor and consumer preference in blackberry.c.1.1 Germplasm and harvest.Overall, 224 fresh-market blackberry samples will be harvested for this objective and evaluated for acids, sugars, volatile aroma attributes, and consumer preference. Five fresh-market blackberry cultivars will be harvested from three different sites across Central and Western Arkansas at two harvest dates in Yr 1 and Yr 2 of the study to investigate the effects of genotype, environment, and genotype × environment interaction on flavor and consumer preference in blackberry. This genotype × environment interaction study will be organized as a three-way factorial with cultivar, site, and year considered fixed effects and harvest date nested within year. The remaining samples evaluated in objective 1 will be unreplicated samples from additional 164 cultivars and breeding selections.Each week during the fall and winter, 16 randomly selected samples will be slightly thawed and pureed using with a Magic Bullet blender (MBR-1101, Los Angeles, CA) within 24 hours prior to evaluation. Aliquots of this pureed sample will then be used for consumer sensory evaluation (400 g), basic composition and organic acids and sugars (200 g), and volatile aroma attribute (200 g) analyses.c.1.2 Consumer sensory analysis. Consumer panels will be conducted weekly during fall and winter (eight during Yr 1, six during Yr 2) at the UA System Sensory Science Center. One hundred consumers will be recruited from a database (N≈6,200) for each weekly panel and selected based on consumption, purchasing habits, and liking of fresh blackberries for each panel.The consumer panels will be conducted as an Augmented Randomized Complete Block Design. All 100 consumers in each of the 14 panels will be served pureed samples of the 'Ouachita' and 'Sweet-Ark® Ponca' checks collected on the same harvest date during Yr 1. The participants in each panel will then be randomly assigned to one of four blocks. In each of these blocks the 25 consumers will evaluate four experimental entries in addition to the 'Ouachita' and 'Sweet-Ark® Ponca' checks. Thus, 16 experimental entries will be evaluated by 25 consumers and the two checks will be evaluated by 100 consumers in each panel.Each consumer will be asked to evaluate overall liking, aroma, flavor, sweetness, and sourness on the 9-point verbal hedonic scale , and blackberry aroma, flavor, sweetness, and sourness on a 5-point JAR scale .c.1.3 Basic Composition. A 200 g aliquot of each pureed sample will be used to determine basic composition attributes (soluble solids, pH, and titratable acidity).c.1.4 Organic acids and sugars. Organic acids and sugars of each sample will be determined using high performance liquid chromatography (HPLC). The peaks will be quantified using external standard calibration based on peak height estimation with baseline integration. Individual sugars, individual organic acids, total sugars (glucose + fructose), and total organic acids (isocitric + malic acid) will expressed as percent (%).c.1.5 Volatile aroma attributes. A 200 g aliquot of each pureed sample will be evaluated for volatile aroma attributes using a Shimadzu Gas Chromatography Mass Spectrometry GCMS-TQ8050 with AOC-6000, and Phaser.A targeted and quantitative approach will be applied the volatiles based on methods described by (Gilbert et al., 2015). Initially, approximately 50 blackberries genotypes will be screened and target volatiles will be identified by comparing collected mass spectra with the spectral library, literature data, and retention indices. Similar to past studies (e.g. Gilbert et al., 2015, Table 1), key volatiles will be selected to represent a number of different chemical classes based on this initial survey. Analytical standards will be purchased for these key volatiles to confirm their identity and quantitative methods will be developed so that the concentrations of each volatile can expressed as µg (volatile)/kg (blackberry) in all of the blackberries analyzed throughout objectives 1 and 2.c.1.6 Data analysis. We will construct genotypic relatedness matrices of the samples in this study using pedigree records and biochemical profiles. A pairwise correlation matrix will be calculated for the sensory and metabolomic data generated in c.1.2-c.1.5. Partial least squares (PLS) analysis will be used to determine the major biochemical determinants of sensory ratings, especially overall liking. Each variable will be centered and scaled and a 10-fold cross validation method will be used. We will also estimate the percent variance explained by sugars, acids, and five to six volatile chemical classes for the nine sensory attributes evaluated by consumers.Objective 2: Conduct genome-wide association mapping of organic acids, sugars, and volatile compounds in blackberry selections from the UA breeding program.c.2.1 Germplasm, harvest, and phenotyping. All of the 224 samples (12 replicate samples of five cultivars for genotype × environment analysis and 164 unreplicated samples from breeding selections and cultivars) will be used in this objective. In addition, we propose to harvest samples from an additional 76 breeding selections in Yr 2 of the project and phenotype them for basic composition, individual sugars and acids, and VOCs following the methods outlined in Objective 1.c.2.2. Genotyping with Capture-Seq. All 245 breeding selections and cultivars evaluated in Objectives 1 and 2 of this proposal will be genotyped with RAPiD GenomicsCapture-Seq technology. The same custom set of 35,054 biotinylated 120-mer Capture-Seq probes developed in USDA NIFA AFRI Plant Breeding for Agricultural Production Program Award # 2018-06274 will be used for this project. DNA libraries will be prepared and enriched using the targeted probe set in the RAPiD Genomics automated laboratory and sequenced with Illumina HiSeq 2500 to achieve at least 80x coverage per marker on average.c.2.3 Data analysis. R Genome wide association analysis will be performed in GWASpoly (Rosyara et al., 2016). The discriminant analysis of principal components (DAPC) technique in R package adegenet (Jombart et al., 2010) will be used to generate the QDAPC matrix representing fixed effects of subpopulations to be used as a covariate for analysis. The random effect kinship matrix will be constructed using the leave-one-chromosome-out (LOCO) method. Additive, simplex dominant, duplex dominant, and diploidized additive models specific to biallelic SNPs in tetraploids will be employed in addition to the general model.Objective 3: Compare genomic and metabolomic selection models for predicting flavor and consumer preference in blackberry.c.3.1 Genomic selection. We will use the Capture-Seq genotype information obtained in c.2.2 to calculate the realized relationship matrix (G) (VanRaden, 2008) using the R package AGHmatrix (Amadeu et al., 2016). AGHmatrix will also be used to calculate the pedigree-based kinship matrix (A). The two matrices will be combined to form a single matrix hybrid matrix (H) that will be used to fit the HBLUP model (Sood et al., 2020). The accuracy of the genomic selection models will be accessed through 10-fold cross-validation using ASReml-r (Butler et al., 2017).c.3.2 Metabolomic selectionWe will obtain a relationship matrix with the metabolomic data (M; including basic composition, acids, sugars, and volatiles measured in objective 1) as described in (Morota and Gianola, 2014). We will assess MBLUPefficiency in predicting the nine sensorial traits evaluated in objective 1. We will also implement XGBoost, which was identified in Colantonio et al. (2022) as the best prediction model for blueberry and tomato.