Predictive Breeding for Wine Quality: From Sensory Traits to Grapevine Genome

Abstract

New pathogen resistant varieties allow an efficient and greatly reduced use of fungicides. These new varieties promise, therefore, an enormous potential to reach the European Green Deal aim of a 50% reduction of pesticides in EU agriculture by 2030. The selection process, and particularly quality evaluation of the wines produced, are a bottleneck slowing down the breeding of new pathogen resistant grapevine varieties. Our major aim is therefore the development of predictive models for wine quality traits. Their implementation in the selection process would considerably increase the efficacy of grapevine breeding.The centrepiece of our study is a segregating white wine F1-population of ‘Calardis Musqué’ and ‘Villard Blanc’ consisting of 150 genotypes with 13 plants per genotype at two locations. A ‘Genotyping by Sequencing’ approach with a novel bioinformatics pipeline delivered a high-density genetic map of the breeding population. Experimental winemaking in a 4-liter scale (micro-vinification) provided authentic wines for comprehensive sensory evaluation and chemical analysis of major and minor metabolites including aroma compounds such as monoterpenoids. Moreover, five annual repetitions at two locations allow robust modelling and an estimation of environmental impact on the phenotypic data. Genetic, metabolic, and sensory data for multiple vintages combine into a comprehensive data base for predictive modelling. The descriptive and quality score card was adapted to the large number of wine samples and the unusual broad range of wine qualities resulting from an unselected set of grapevine genotypes. Based on evaluation of all 150 genotypes we differentiated a set of best and worst wines reproducibly over years. Environmental-related differences among vintages were still present. Intensity of the descriptive attribute “floral” played a crucial role for total quality within this population and correlates with linalool and cis-rose oxide concentration of the wines in all vintages measured by SIDA-SPE-GC-MS. In addition, total concentrations of linalool enabled the discovery of several genomic regions (quantitative trait loci, QTLs) that collocate with putative genes associated with terpene biosynthesis. Multi seasonal data allowed refinement and validation of models predicting these wine quality traits. Further exploitation of the large data set will provide more insights into genomic regions related to other wine quality traits and will allow an early selection of genotypes of promising genetic quality potential or sorting out of poor candidates during grape vine breeding.