**1. Introduction**

Plant species exhibit a high ability for local adaptation and phenotypic plasticity that may contribute to their distribution range. While local adaptation is the genetically fixed advantage of a population under certain environmental conditions [1], phenotypic plasticity is the ability of a genotype to generate different phenotypes in response to variation in the environment [2,3]. This variation is created by mutation, recombination, and introgression, and by population genetics processes, such as genetic drift and natural selection, that determine its evolutionary fate. Understanding of the genetic

basis of local adaptation and phenotype plasticity is relevant to climate change, crop production, conservation, and understanding of speciation. The combination of genomics and ecology enables genome-wide analysis to reveal the interaction between organisms and environment [4] and to identify genomic regions involved in adaptation [5]. On the other hand, phenotypic plasticity may allow species to grow and survive in di fferent environments despite a restricted genetic base. Thus, phenotypic plasticity could be advantageous under variable environments, including climatic change [6], and may also increase species invasion success [7].

Timing of seed germination is one of the key steps in plant life, influencing the subsequent destiny of individuals as well as whole populations at determined area. Plants have evolved various mechanisms to control seed germination within- and among-seasons and in relation to the diversity of climates, habitats, and biotic pressures [8,9]. Three di fferent kinds of dormancy have been described to allow optimal germination timing under specific environmental conditions [8,9]: (1) morphological dormancy (MD) refers to seeds that have an underdeveloped embryo and require time to grow; (2) physiological dormancy (PD) prevents embryo growth and seed germination until chemical changes occur, involving abscisic acid and gibberellins metabolism, among other factors; and (3) physical dormancy (PY) is caused by water-impermeable palisade cells in the seed coat. PY occurs in at least 18 Angiosperm families and is frequent in legumes [10–12].

Adaptation to the local environment operates through selection for successful germination and early plant establishment [13]. The prevention of germination of a certain proportion of seeds even under optimal conditions for germination reduces the risk of mortality in less predictable environmental conditions. It has been suggested theoretically [14] and shown empirically [15] that adaptation for dormancy is a bet-hedging strategy to magnify the evolutionary e ffect of "good" years and to dampen the e ffect of "bad" years, i.e., to bu ffer environmental variability [16]. In addition, species that frequently and reliably produce seed can a fford riskier germination under unfavorable conditions (e.g., small rainfall events) because the consequences of failure to establish are less dire than for species that do not reliably produce seed [17]. Desert annuals that do not frequently and reliably reproduce are model organisms for the study of the bet-hedging strategy [18].

In order to germinate, specific environmental conditions need to be met to break the seed dormancy [8]. However, less is known about the factors which release the PY dormancy. Through experimental studies, it was shown that, in addition to scarification, wet or dry heat were found to be e ffective [8,10,19]. In addition, natural conditions, such as temperature and soil moisture oscillations, are the major players [20,21]. Laboratory studies have demonstrated an association between seed responsiveness to temperature and environmental thermic characteristics [10,22,23]. However, only limited data are available on how and why PY varies inter- and intra-specifically in natural ecosystems [24]. Legumes are thus a model example for studies of PY dormancy patterns in relation to environmental variations. The study of Rubio de Casas et al. [25] showed a latitudinal gradient in PY dormancy in legumes. Thus, PY dormancy increases from regions with long growing seasons (e.g., tropical climate) in lower latitudes to regions with a seasonal climate in higher latitudes. However, there are some studies of intraspecific PY dormancy variation along environmental gradients in several legume species [10,26–29] that are in disagreement with the results of Rubio de Casas et al. [25]. *Medicago truncatula* (barrel medic) is an annual, diploid, self-fertile species with a natural geographic distribution across the Mediterranean Basin. Phenotypic variation among populations has been explained by the adaptation to local environmental conditions [30]. *M. truncatula* offers an excellent model to study seed dormancy in relation to genetic and environmental factors because within its range it inhabits environments with rather contrasting climatic conditions, di ffering not only in mean annual temperature and precipitation, but also in within- and across-season variability (unpredictability). Its seeds exhibit both physical and physiological dormancy. Physiological dormancy in *M. truncatula* seeds is non-deep, and is removed during the seed ripening period [31,32]. The short after-ripening period to overcome PD (<3 months) determines that PY release is the most important trait to regulate the timing of seedling emergence. Despite this, most germination studies in *M. truncatula* eliminate

the influence of PY dormancy through prolonged periods of storage (>9 months) and/or by seed scarification [33,34]. Large georeferenced collections, a reference genome, and a high-density single nucleotide polymorphism (SNP) map of more than 260 genotypes of *M. truncatula* are available [30,35] and were used for study of the association between the genome and the environment in relation to flowering [35–37]. We have taken advantage of this georeferenced collection to analyze the patterns of dormancy release in 178 accessions of *M. truncatula* originating from various environments in the Mediterranean basis and tested seeds under alternating temperatures. The following questions were addressed: (i) Is there variation in physical seed dormancy among accessions to temperature treatments? (ii) Which of the ecological factors acting as potential adaptation drivers are correlated with dormancy? (iii) Are there any candidate genes that might be related to seed dormancy release in *Medicago*, using genome-wide association (GWAS) analysis?

Our study showed that phenotypic plasticity of final dormancy was significantly correlated with increased aridity, suggesting that plastic responses to external stimuli provide seeds with strong bet-hedging capacity and the potential to cope with high levels of environmental heterogeneity. Genome-wide association analysis performed on seven seed dormancy traits and three bioclimatic variables identified 136 candidate genes as potential regulators of physical dormancy. A large proportion of candidate genes were annotated as involved in synthesis of secondary metabolites, in cell wall modification, and hormone regulation. The knowledge about the regulation of seed dormancy by environmental factors could be extended to other legume species, particularly to crop wild relatives of economically important crops, such as chickpea, lentil, and faba bean. In addition, it can be used in a conservation biology context for the managemen<sup>t</sup> of endangered plant species in relation to climate change.
