3.2 Epigenetic changes in response to Cd and Cu
Overall, DNA methylation in S. cataractae was very low (Table 1). For the complete SMP matrix (n=43,365 SMPs) mean cytosine methylation per group averaged 3.3% ± 0.06 in all contexts considered together, and 0.69% ± 0.05 in CG, 12% ± 0.05 in CHG, and 0.71% ± 0.07 in CHH. Most of the cytosines in the CG, CHG, and CHH contexts were unmethylated, i.e. methylation level ≤ 5% in ≥ 95% of the samples (92.6% in CG, 91.2% in CHH and 81.9% in CHG); 1.9% of the cytosines in the CHG context were fully methylated, i.e. methylation level ≥ 95% in ≥ 95% of the samples (Fig. S2A). The standard deviation of DNA methylation averaged 16.0% ± 0.07 in all contexts together, and 3.1% ± 0.39 in CG, 31.5% ± 0.05 in CHG, and 2.8% ± 0.39 in CHH (Table 1). For the complete and polymorphic SMP matrix (n=3,769 SMPs), mean cytosine methylation per group averaged 10.8% ± 0.25 in all contexts, and 2.8% ± 0.26 in CG, 34.3% ± 0.23 in CHG, and 2.4% ± 0.32 in CHH (Fig. S2B). The standard deviation of DNA methylation averaged 27.5% ± 0.15 in all contexts, and 7.2% ± 0.88 in CG, 45.3% ± 0.18 in CHG, and 5.3% ± 0.86 in CHH (Table 1).
DNA methylation levels differed among genomic features with repeats generally showing higher methylation levels than genes and transposons; for example, mean methylation across samples in repeats was 0.9, 12.6, and 1% in the CG, CHG, and CHH contexts respectively compared to the ⁓0.5, 10.2, and 0.6% in the CG, CHG, and CHH contexts found for genes and transposons respectively (Fig. S3).
The dbRDA performed to test the effect of population, treatment, and their interaction on genome wide DNA methylation in S. cataractaewas not significant when run with the epigenetic distance matrix calculated comparing the methylation level only (Table S4). When run with the distance matrix obtained comparing both methylation level and methylation variance between samples, the model was significant for both common garden experiments, and explained 61% and 56% of the variation of the epigenetic distances in the Cd vs. C and the Cu vs. C comparison respectively (Table 2; results for each separate sequence context were similar and are shown in Table S4). The effect of both predictors and their interaction was significant in both experiments: treatment explained ~9% and 7% of the variation of the epigenetic distances in the Cd vs. C and Cu vs. C comparisons respectively; population explained ~21% of the variation in both experiments; and the interaction explained 33% and 29% of the variation in the Cd vs. C and Cu vs. C comparisons respectively (Table 2). Both metal treatments induced more DNA methylation variability among samples of Sc2 (one of the most metal tolerant populations); this effect was weaker, yet noticeable on Sc1 and Sc2 in response to Cu, and on Sc4 in response to Cd (Fig. 1A1, A2). The first two constrained axes of the dbRDA explained ~35% and ~29% of the variation in epigenetic distances in the Cd vs. C and the Cu vs. C comparison respectively. Most of the variation in the first axis (RDA1) was driven by the differentiation of treated samples from Sc2 in both common garden experiments, whereas most of the variation in the second axis (RDA2) was driven by the differentiation of control samples from Sc3 and Sc4 in the two common garden experiments (Fig.1A1, A2).
We found between 25-78 differentially methylated positions (DMPs) between Cd-treated and control (Fig. 1B1) and Cu-treated and control plants (Fig. 1B2) across populations (between 0.66-2.1% of the polymorphic cytosines; Fig. S4). All populations showed more DMPs within the CHH context (17%, 26%, and 57% of all DMPs in CG, CHG, and CHH contexts respectively; Table S5) but the proportion of DMPs per context did not differ from the proportion of cytosines interrogated per context. Sc3 showed the highest number of DMPs in both treatments (Figs. 1B1, 1B2, S4). Cadmium induced more DMPs than Cu (except in Sc4; Figs. 1B1, 1B2, S4). We found more DMPs with a negative methylation change, i.e. hypomethylation, in Sc2 in response to Cd and Cu, and in Sc3 in response to Cu; Sc1 and Sc4 showed more DMPs with a positive methylation change, i.e. hypermethylation, in response to Cd and Cu respectively. Most DMPs were unique for each population except four: two shared between Sc1 and Sc2 and one shared between Sc1 and Sc4 in the Cd vs. C comparison; one shared between Sc3 and Sc4 in the Cu vs. C comparison (Fig. S5A, S5B). The two metals, however, induced a considerable number of common DMPs within each population, most of which were hypomethylated: 29, 49, 41, and 12% of the DMPs were common to both treatments in Sc1 to Sc4 respectively (Fig. S5C, underlined DMPs in Table S5).