Adaptive significance of flowering time plasticity: synthesising 10 years of Arabidopsis research in the field.
Authors: Alexandre Fournier-Level 1; Andhika R Putra 1; Johanna Schmitt 2
Affiliations: 1) The University of Melbourne; 2) UC Davis
Keywords: Ecological & conservation genetics
Genetic and environmental variation combine to promote trait plasticity. Despite well-laid theoretical expectations, empirical data are still indispensable to validate the importance of plasticity for fitness and evolution in natural conditions. Here we synthesise 10 years of field experiments on Arabidopsis thaliana’s flowering time where hundreds of natural accessions were planted across multiple European sites and seasons.
Extensive flowering time plasticity was observed, primarily across environment, but also between accessions. The pattern of plasticity was mostly driven by ecotypes originating from marginal regions (Nordic and high elevation) and expressed over the summer, not the most common growing season for this species. Selection analysis showed that highly plastic genotypes were negatively selected in most cases, suggesting that plasticity was primarily cryptic and expressed away from home conditions. However even if selection primarily favoured highly canalized, early flowering, we did observe the selection of genotypes with increased seasonal plasticity, able to delay flowering in favourable Mediterranean fall conditions while still flowering early in the spring. Genome-wide association study identified GIGANTEA SUPPRESSOR 5 as main candidate gene for this plastic response, consistent with previous lab observations. Nonetheless, the massive gap between the overall very high heritability for flowering time plasticity traits and the low amount of variance explained by individual loci underscores a very polygenic architecture.
The data collected were used to train a genomic prediction model of flowering time across Europe. We show that a LASSO-penalized linear-mixed model designed using daily minimum and maximum ground temperature and a genetic kinship matrix without individual locus effects performed well with a cross-validated accuracy of 0.93. We also showed that a genotype-by-environment effect substantially improved the predictions, emphasizing the importance of capturing plasticity. This predictive model was independently validated using data from Rhode Island showing a good transferable accuracy of 0.64. These predictions were used to identify genotypes that could be used to genetically offset the predicted consequence of climate change in an environment-specific manner. We highlight the importance of multi-site testing of large numbers of genotypes from multiple origins to best design climate change-ready restoration using genomic data.