Epigenetic change is considered relatively unstable and short-lived, questioning its contribution to long-term adaptive potential. However, epigenetic modifications can accumulate in the presence of environmental stress, resulting in beneficial epigenetic memories where environments are challenging. Diverging epigenetic memories have been observed across large spatial scales, and can persist through multiple generations even in the absence of the causative environmental stressor. It is unknown, however, to what extent epigenetic variation contributes to fine-scale population structure and evolution. We compared DNA methylation patterns between a steep, altitudinal gradient (<2 km) and a wide spatial gradient (>500 km) using whole genome bisulfite sequencing data from 30 Fragaria vesca plants germinated and grown in controlled conditions. To assess the stability of spatial epigenetic variation in the presence of an environmental stressor, we applied acute drought stress to part of the plants and quantified drought-induced changes in DNA methylation signatures. We find that epigenetic memories and genomic islands of epigenetic divergence arise even at fine spatial scale, and that distinct spatial scales are featured by distinct epigenetic patterns. For example, demethylation of transposable elements consistently occurred at the large but not the fine spatial scale, while methylation differentiation for most biological processes were shared between spatial scales. Acute drought stress did not result in significant epigenetic differentiation. Our results indicate that population history, rather than short-term environmental stress, plays a dominant role in shaping epigenetic signatures. Specifically, repeated historical stress levels associated with heterogeneous environmental conditions may be required for acquiring a stable epigenetic memory and for coping with future environmental change.

Whole genome bisulfite sequencing was used to identify methylated cytosines across 30 genomes of wild woodland strawberries raised in a common garden. The seeds were collected in France (west and east) and Poland (west). Differentially methylated cytosines (DMCs) were called using the R package methylKit version 1.10.0 with conventional DMC parameters (Akalin et al. 2012). Specifically, only cytosines with at least 5x coverage in at least 3 samples per group were retained (Walker et al. 2015; Wan et al. 2016). To reduce bias due to outlier depth, bases with a read depth above the 99.9th percentile of coverage are filtered out. The filtered data were used to test for differentially methylated cytosines (DMCs), considering a 25% difference and q-values <0.01 as significant. DMCs were identified between (i) low, mid and high altitudinal samples (hereafter “altitudinal DMCs”), (ii) three distance European samples (hereafter “spatial DMCs”), and (iii) two soil moisture treatments (hereafter “drought DMCs”).

The dataset includes statistically significant DMCs (differentially methylated cytosines) along a fine-scale altitudinal gradient, a large-scale spatial gradient, and between soil moisture treatments, along with methylation proportions at each sampled location, and whether or not the genes alining with the F. vesca genome were involved with transposon activity.