Experimental design
The MAGIC population was grown in three different geographical regions (France, Israel and Morocco) and four specific stress treatments were applied. Trials were conducted in order that in a given trial any stress treatment was applied aside a control trial (Supplemental Table 1). Treatments consisted in water deficit (WD), two levels of salinity – considered here as low salinity (LS) and high salinity (HS) – and high temperature (HT) stress. Water deficit was applied by reducing the water irrigation of about 70% and 30% according to the reference evapotranspiration in Israel in 2014 and 2015, respectively and by 50% in Morocco in 2015. Salinity treatment was managed as described in Diouf et al. (2018) and the average electrical conductivity of the substrate (Ec) in Morocco 2016 was 3.76 and 6.50 dS.m-1 for LS and HS, respectively; while the Ec in the control condition in Morocco 2015 was about 1.79 dS.m-1. For HT stress, plants were sown during the late spring and phenotyped in the summer 2014 in Israel (HIs14) and summer 2017 in France (HAvi17). During HT treatments, greenhouse vent opening was managed all along the entire growing season, with opening the vent only when temperatures rose up to 25°C. Average mean (respectively maximal) temperatures calculated on daily (24 hours) measurements were of 26°C (respectively 34°C) for HAvi17 and 33°C (respectively 48°C) for HIs14. Besides stress treatments, local conventional cultural conditions were applied for control treatments as described in Diouf et al., (2018).
Environments were considered as any combination of a geographical region, a year of trial and an applied treatment (Supplemental Table 1). Climatic sensors were installed in the greenhouses and climatic parameters recorded hourly in all environments. From the climatic parameters, seven environmental covariates were defined (Supplemental Figure 1) including temperature parameters (mean, minimal and maximal daily temperatures and thermal amplitude), the sum of degree-day (SDD), the vapour-pressure-deficit (Vpd in kPa) and the relative humidity (RH) within the greenhouse. To characterize the environments, every covariate was calculated during the period covering flowering time of the population on the fourth truss. Indeed, phenotypic data analyzed here were mostly recorded on the fourth and fifth trusses (Supplemental Table 2). Hierarchical clustering was performed with ‘FactoMineR’ R package (Lê et al., 2008) using the environmental parameters to group environments according to their similarity regarding the within-greenhouse climatic conditions.
The MAGIC population, the eight parental lines and the four first generation hybrids (one hybrid per two-way cross) were evaluated for fruit weight (FW) by measuring the average FW of the third and/or fourth plant truss in each environment. Phenotypic data were recorded across the different environments for nine supplemental traits related to fruit quality – fruit fruit firmness (firm) and soluble solid content (SSC); plant phenology – flowering time (flw), number of flowers (nflw) and fruit setting (fset); plant development – stem diameter (diam), leaf length (leaf) and plant height (height) and fruit number (nfr). Details about the phenotyping measurements are in Supplemental Table 2. At least two plants per MAGIC line were replicated in each environment except in Avi17 (control condition) where the average phenotype was recorded from single plant measurements. Parents and hybrids had more replicates per genotype (at least two) and served as control lines to measure within-environment heterogeneity.