Introduction
The timing of seed germination affects fitness by determining conditions during seedling establishment and growth (Donohue et al. 2010; Postma and Ågren 2016), but may also influence fitness by having cascading effects on later life-history traits such as flowering start (Evans and Cabin 1995; Donohue 2002; Wilczek et al. 2009; Akiyama and Ågren 2014; Burghardt et al. 2015; Martínez-Berdeja et al. 2020; Zacchello et al. 2020). In seasonal environments, there should be strong selection for timing of germination matching periods favourable for seedling establishment and growth (Donohueet al. 2010; Wadgymar et al. 2015), and adjustment of germination timing is expected to critically affect the ability of species to adapt to climate change (Fenner and Thompson 2005; Cochraneet al. 2015). The degree to which among-population variation in seed dormancy is related to climatic differences across different spatial scales thus becomes of considerable interest for assessing the potential of adaptive evolution in response to climate change (Dawsonet al. 2011).
In many species, timing of germination is regulated by the strength of seed dormancy. Primary seed dormancy, i.e., dormancy at the time of seed maturation, and the rate at which it is lost will determine when seed germination will be triggered in response to environmental cues (Vleeshouwers et al. 1995; Li and Foley 1997; Finch-Savage and Leubner-Metzger 2006; Bewley et al. 2013; Baskin and Baskin 2014). Optimal seed dormancy should be positively correlated with the length of the period following seed release that is unfavourable for seedling establishment (Meyer and Monsen 1991; Allen and Meyer 1998; Llorens et al. 2008; Wagmann et al. 2012), and may thus vary among populations. Genetic variation in seed dormancy has been documented among populations of several species and across various spatial scales, for example in Arabidopsis thaliana within Europe (Kronholm et al. 2012; Debieu et al. 2013), inBromus tectorum in North America (Allen and Meyer 2002), and inDigitaria melanjiana in central and eastern Africa (Hacker 1984). Several environmental factors may drive divergence of seed dormancy, and one approach to identify potential agents of selection is to examine the correlation between phenotype and environment at the site of origin (Wadgymar et al. 2017). Characterizing spatial variation in seed dormancy and its association with differences in environmental factors can thus provide an insight to the possible drivers of such variation.
Among-population variation in seed dormancy can be the result not only of genetic differentiation, but also of environmental effects and the interaction between these two factors (Young et al. 1991; Benderet al. 2003; Schütz and Rave 2003; Donohue et al. 2005; Donohue 2009; Postma and Ågren 2015; Fenner 2018). Most studies that document within-species variation in seed dormancy have grown populations in the greenhouse (e.g., Allen & Meyer, 2002; Debieu et al., 2013; Kronholm et al., 2012; Vidigal et al., 2016; Wagmann et al., 2012). However, plants raised in the greenhouse are typically exposed to temperature, light and watering regimes that are very different from those experienced in natural populations. Studies comparing dormancy of seeds produced in the greenhouse and in the field have also detected strong genotype × maternal environment interactions (Schütz and Rave 2003; Fernández-Pascual et al. 2013; Postma and Ågren 2015). To determine the importance of genetic differentiation, environmental effects and their interaction for variation in seed dormancy, seeds of different source populations should thus ideally be produced in multiple relevant field environments (Young et al. 1991; Schütz and Milberg 1997).
In this study, we quantify variation in primary seed dormancy among populations of the annual model organism A. thaliana in two regions in Europe and examine the association between within-region variation in seed dormancy and climate. A. thaliana is native to Africa and Eurasia (Durvasula et al. 2017), and occurs in habitats that vary widely in the length of the dry summer period. In winter-annual populations of A. thaliana , seeds are matured in late spring and early summer and germinate in autumn (Ågren and Schemske 2012). As in other species with physiological dormancy, germination inA. thaliana cannot occur until seed dormancy has been released by a process called after-ripening (Finch-Savage and Leubner-Metzger 2006; Montesinos-Navarro et al. 2012). Considerable within-species variation in the required length of after-ripening has been observed at different spatial scales, but its association with climatic factors is still unclear. Kronholm et al. (2012) found a negative association between seed dormancy and summer precipitation among European populations, and Vidigal et al. (2016) similarly found that high dormancy was associated with high temperature and low summer precipitation within the Iberic peninsula. By contrast, Debieu et al. (2013) found no climatic association but only a latitudinal cline across Europe. Moreover, in a study of regional variation in north-eastern Spain, high primary dormancy was associated with dry, but cold environments of origin (Montesinos-Navarro et al. 2012), indicating that large-scale climatic associations are not necessarily reflected at smaller spatial scales. Additional studies of regional variation are therefore needed to characterize spatial patterns of seed dormancy differentiation in A. thaliana . Furthermore, seed dormancy in A. thaliana is strongly affected by the maternal environment and by the interaction between genotype and maternal environment (Postma & Ågren, 2015), suggesting that the possibility of such interactions should be considered when examining correlations between trait expression and environment at site of origin.
Here, we planted individuals from 28 Fennoscandian populations and 17 Italian populations (4-5 maternal lines per population for a total of 224 maternal lines) in three common gardens, one at the site of a natural population in north-central Sweden, one at the site of a natural population in central Italy, and one in a greenhouse in Uppsala, Sweden (Fig. 1 ). Fennoscandia and Italy represent the northern and southern range margins of A. thaliana in Europe. The summer period during which intermittent droughts is likely to kill any young emerging seedlings is longer in Italy compared to Fennoscandia (Fig. 2A and B ). Consequently, we expect Italian populations to show stronger seed dormancy than Fennoscandian populations, as previously observed in a comparison between the populations native to the Swedish and Italian field sites used in the present study (Postma and Ågren 2015). Similarly, within regions, seed dormancy can be expected to be positively correlated with the length of the period with summer conditions and with temperature during this period, and negatively correlated with precipitation during summer.
We tested the hypotheses that (i) dormancy of seeds produced by Italian populations is stronger than that of seeds produced by Fennoscandian populations, (ii) seed dormancy is affected by the maternal environment and its interaction with region and population of origin, and (iii) within regions, seed dormancy is related to climatic conditions at the site of origin in summer.