Discussion
This study has shown that across three different seed production environments, Fennoscandian and Italian Arabidopsis thalianapopulations differ consistently in seed dormancy. As expected from differences in summer temperature and precipitation, Italian populations produced seeds with stronger dormancy than did Fennoscandian populations. In addition, we found wide variation in dormancy among populations within each of the two geographic regions, and that for most populations seed maturation environment had a strong effect on seed dormancy. Below we discuss the results in relation to previous studies documenting variation in seed dormancy among and within geographic regions, effects of seed maturation environment on dormancy level, and processes affecting population differentiation in this trait.
The stronger dormancy of seeds produced by Italian compared to Fennoscandian populations is in accordance with expectations based on climatic differences between the two regions. In the Italian populations sampled, seed maturation in late April is followed by a long, hot and dry summer, during which germination in response to occasional rain is bound to be associated with high seedling mortality. By comparison, the length of the period after seed maturation that is unfavourable for seedling establishment is markedly shorter in Fennoscandian populations. As a result, selection is expected to favour stronger primary seed dormancy in the Italian compared to the Fennoscandian populations (cf. Postma & Ågren, 2016). In experiments conducted at the two field sites, August was identified as the optimal time of germination at the Swedish site (Akiyama and Ågren 2014), and November at the Italian site (Zacchello et al. 2020). The difference in seed dormancy between Italian and Fennoscandian populations documented in the present study is consistent with previous observations indicating a decrease in seed dormancy with increasing latitude of origin among A. thalianaaccessions sampled across Europe (Kronholm et al. 2012; Debieuet al. 2013), and is likely to be representative for differences between north European populations and southern populations at low altitude in general.
Seed dormancy was strongly affected not only by the region of origin but also by the maternal environment. Seed dormancy after 12 weeks of after-ripening was stronger among seeds produced at the Swedish field site than among seeds produced at the Italian field site, which is consistent with differences observed in a former study documenting seed dormancy of a population of recombinant inbred lines (RILs) planted at the two sites and in the greenhouse (Postma and Ågren, 2015). Differences in temperature during seed maturation may have contributed to the observed difference in seed dormancy between the two field sites. Low temperature during seed maturation has been found to increase seed dormancy of A. thaliana (Chiang et al. 2011; Footittet al. 2011, 2013; Kendall and Penfield 2012; He et al.2014; Coughlan et al. 2017; Kerdaffrec and Nordborg 2017), but also in a wide range of other species including Avena fatua ,Beta vulgaris , Chenopodium bonus-henricus , andPlantago lanceolata (see review Fenner 2018). During the two months preceding seed dispersal (i.e., March and April in Italy, and May and June in Sweden), air temperature was 1.5 ˚C colder in Sweden than in Italy (Italy: 12.7˚C; Sweden: 11.2˚C; data recorded at the sites using loggers as in Ågren & Schemske, 2012).
More surprising was the low germinability of seeds 12 weeks after harvest in the greenhouse (Fig. 3 ). Because temperature in the greenhouse was higher than at the field sites during seed maturation, we expected seeds produced in the greenhouse to have the lowest dormancy, as observed in the experiment with the RIL population (Postma and Ågren 2015). The present results indicate that environmental effects on development of seed dormancy may vary among experiments also in a greenhouse with a rather well-controlled temperature regime. Further studies are needed to examine the possible influence of differences in soil nutrient concentrations (Baskin and Baskin 2014) and water content (Alboresi et al. 2005) for seed dormancy development in this environment.
Correlations between germination proportions and measures of climate at sites of origin were generally weak, and statistically significant only for seeds produced by Fennoscandian populations in the greenhouse (Table 4 ). One and three weeks after maturation, seed dormancy tended to be negatively related to precipitation and positively related to summer temperature at the sites of origin (Table 3 and 4 ). These correlations are in line with predictions, and with associations between seed dormancy and climate observed within the Iberic peninsula (Vidigalet al. 2016), and at a larger scale across Europe (Kronholmet al. 2012). In contrast to correlations documented for the Fennoscandian populations, primary seed dormancy increased from sites characterized by high temperature and wet conditions to those characterized by lower temperature and dryer conditions among A. thaliana populations sampled along an altitudinal gradient in north-eastern Spain (Montesinos-Navarro et al., 2012). The contrasting results show that correlations between seed dormancy and climatic variables vary among regions, and suggest that the strength and direction of correlations with different climatic variables will depend on which part of the overall climatic variation is examined.
There are several possible reasons for the generally weak correlations between primary seed dormancy and large-scale climatic variation within regions, and the lack of statistically significant associations for Italian populations. First, more populations were sampled in Fennoscandia and the climatic range represented by these populations was wider than that represented by the Italian populations (Table S2 ), which should increase the chance of detecting relationships between seed dormancy and environmental variables. Second, lower survival and fecundity in the field compared to the greenhouse resulted in smaller sample sizes, which should have reduced precision of estimates and statistical power in analyses of variation in dormancy among seeds matured in the field. Third, large-scale climatic data may not well represent local microclimate since the latter is strongly influenced by topography and exposure. For example, most of the northernmost populations grow on steep, south-facing slopes, which represent particularly warm and dry habitats in the landscape. Fourth, in addition to micro-climatic conditions, optimal germination time and seed dormancy may depend on environmental factors, such as soil composition, which affects water-holding capacity, and on vegetation cover, which affects intensity of competitive interactions. Fifth, seed dormancy of present-day populations may not mirror optimal seed dormancy at the sites of origin, but rather reflect founder events or genetic correlations with traits more strongly related to fitness. This may seem less likely considering the strong effects of germination date for likelihood of seedling establishment, survival and fecundity in A. thaliana (e.g., Donohue et al., 2005; Akiyama & Ågren, 2014; Postma & Ågren, 2016; Zacchello et al., 2020). However, germination date is determined not only by dormancy at the time of seed maturation, but also by processes affected by the post-dispersal environment, such as rate at which dormancy is released, and possible acquirement of secondary dormancy (Montesinos et al., 2012; Postma et al., 2016; Martínez-Berdeja et al., 2020). To further explore the consequences of the documented variation in primary dormancy, it would be of interest to compare dormancy release under contrasting field conditions and examine whether any genotype × field environment interaction can be detected in this trait, since this should influence the realized germination time.
In conclusion, this study has documented strong differentiation in seed dormancy between Fennoscandian and Italian populations of A. thaliana, but also among populations within each of the two regions. The wide variation in seed dormancy documented among populations within the two geographic regions indicates considerable evolutionary flexibility, and is consistent with strong divergent selection on this trait. Within Fennoscandia, which was the best sampled region, we found an association between seed dormancy and temperature and precipitation, two climatic factors that are expected to change in the future (Masson-Delmotte et al. 2018). Reciprocal seed and seedling transplants could be used to determine whether this among-population differentiation in seed dormancy contributes to local adaptation. Moreover, to assess the potential for adaptive evolution in response to ongoing changes in climate, future studies should examine the extent to which seed dormancy varies genetically within natural populations, and whether current gene flow among divergent populations is sufficient to maintain such variation.