INTRODUCTION
Plants are sessile organisms which have to cope with environmental fluctuations to ensure species reproduction for persistence in nature. For a given genotype, the expression of different phenotypes according to the growing environment is commonly called phenotypic plasticity (PP) (Bradshaw, 1965). It offers the possibility to plants to adapt to new environments, notably new locations, changes in climatic conditions or seasonal variations. In agriculture, the range of environmental variation for crop cultivation may also include different cultural practices or growing conditions, leading to the expression of PP on agronomic traits and unstable performance. When different genotypes/accessions are examined for PP within a species, inter-individual variations in their responses usually lead to the common phenomenon of genotype-environment (GxE) interaction (El-Soda et al., 2014). Understanding the genetic mechanisms driving PP and GxE in plants is a crucial step for being able to predict yield performance of crop cultivars and to adapt breeding strategies according to the targeted environments.
In plants, the genetic basis of PP has been investigated to assess whether PP has its own genetic regulation and thus could be directly selected. Three main genetic models, widely known as the over-dominance, allelic-sensitivity and gene-regulatory models were proposed in the literature as underlying plant PP (Scheiner, 1993; Via et al., 1995). The over-dominance model suggests that PP is negatively correlated to the number of heterozygous loci (Gillespie and Turelli, 1989). The heterozygous status is favored by allele’s complementarity in this case. Allelic-sensitivity and gene-regulatory models are assumed to arise from the differential expression of an allele according to the environment and epistatic interactions between structural and regulatory alleles, respectively. The latter assumes an independent genetic control of mean phenotype and plasticity of a trait. Using a wide range of environmental conditions, the prevalence of the allelic-sensitivity or gene-regulatory model in explaining the genetic architecture of PP was explored in different crop species including barley (Lacaze et al. 2009), maize (Gage et al., 2017; Kusmec et al., 2017), soybean (Xavier et al., 2018) and sunflower (Mangin et al., 2017).
Quantification of PP is however a common question when analyzing the genetic architecture of plasticity since different parameters for PP estimation are available as reviewed by Valladares et al. (2006). At a population level, when multiple genotypes are screened in different environments, different approaches can be used to assess plasticity (Laitinen and Nikoloski, 2019). The most common of these approaches is the joint regression model (Finlay and Wilkinson, 1963) that uses the average performance of the set of tested genotypes in each environment as an index on which the individual phenotypes are regressed. This model, commonly known as the Finlay-Wilkinson regression model, allows to estimate linear (slopes) and non-linear plasticity parameters (from the residual errors) that presumably have different genetic basis (Kusmec et al., 2017). If the detailed description of the environments is available, the environmental index used in the Finlay-Wilkinson regression model can be replaced by environmental covariates such as stress indexes through factorial regression models (Malosetti et al. 2013). Thus plasticity could be estimated as the degree of sensitivity to a given stress continuum (Mangin et al., 2017).
Climate change is predicted to increase the frequency and intensity of abiotic stresses with a high and negative impact on crop yield (Zhao et al., 2017). Plants respond to abiotic stresses by altering their morphology and physiology, reallocating the energy for growth to defense against stress (Munns and Gilliham, 2015). Consequences on agronomic performances are apparent and detrimental to productivity. The most common abiotic stresses studied across species are water deficit (WD), salinity stress (SS) and high temperature stress (HT). The negative impact of these stresses on yield have been underlined for major cultivated crops; however, positive effects of WD and SS on fruit quality have been observed in fruit trees and some vegetables notably in tomato (Costa et al. 2007; Mitchell et al. 1991; Ripoll et al. 2014).
Tomato is an economically important crop and a plant model species which led to numerous studies that contributed much in understanding the genetic architecture of the crop and its response to environmental variation. However, most of the studies that addressed the genetic architecture of tomato response to environment were conducted on experimental populations exposed to two conditions (i.e. controlvs stress). Albert et al. (2018) for example identified different WD-response quantitative trait loci (QTL) in a bi-parental population derived from a cross of large and cherry tomato accessions. Tomato heat-response QTLs were also identified in different experimental populations including interspecific and intraspecific populations (Grilli et al., 2007; Xu et al., 2017a; Driedonks et al., 2018). These studies investigated heat-response QTLs using mostly reproductive traits screened under heat stress condition. Villalta et al. (2007) and Diouf et al. (2018) investigated the genetic architecture of tomato response to SS and identified different QTLs for physiological and agronomic traits, involved in salinity tolerance. However, no QTL study has yet been conducted on tomato plasticity assessed under a multiple stress design, although the coincidence of different stresses is a more realistic scenario in crop cultivation, especially with the climate change.
Tomato benefits of a large panel of genetic resources that have been used in multiple genetic mapping analyses (Grandillo et al. 2013). Bi-parental populations were first used in QTL mapping and permitted the characterization of plenty of QTLs related to yield, disease resistance and fruit quality. In the genomic era, new experimental populations were developed offering higher power and advantages for QTL detection. These include mutant collections, BIL-populations and multi-parent advanced generation intercross (MAGIC) as described in Rothan et al. (2019). The first tomato MAGIC population was developed at INRA-Avignon in France and is composed of about 400 lines derived from an 8-way cross (Pascual et al. 2015). This population showed a wide intra-specific genetic variation under control and stress environments and is highly suitable for mapping QTLs (Diouf et al., 2018).
In the present study, we used the 8-way tomato MAGIC population described above and evaluated its response in a multi-environment trial (MET) design. The population was grown in 12 environments including control and several stress conditions (WD, SS and HT), and agronomic traits related to yield, fruit quality, plant growth and phenology were measured. Different plasticity parameters were computed and used together with mean phenotypes to decipher the genetic control of response to environmental variation. Multi-environment QTL analysis was performed in addition to detection of interactive QTLs (QEI) along with QTL mapping for plasticity traits.