Supplementary Figure 8: Percentage of positions identified as DMPs in CHG and CG contexts.

The upper panel shows the percentage of cytosines in CHG context detected as DMP, whereas the lower panel shows the same for cytosines in CG context. Results are split for gene related and non-gene related clusters. Data for all positions per species is added as an excel file (Supplementary data 2). DMP detection was carried out as described in online methods section “DMP detection in all species”. For Arabidopsis we found that 11% of cytosines in CG context were differentially methylated in genes and 8% in non-genes. These values are lower than the 17.8% of cytosines in CG context reported as DMPs in Becker et al 2011, We suspect that our values are lower due to the stringent selection criteria requiring methylation to be present in both forward and reverse reads on both Watson and Crick strand. This might lead to under-estimation of actual DMP levels at lower sequencing coverage. Indeed, by selecting a higher minimum read coverage threshold of 50 forward and 50 reverse informative reads the percentage of CG dinucleotides found in genes was higher with 14.5 % and in non- genes 11.3 %. For most species, CG-DMPs are more prevalent in genes, with the exception of Scabiosa columbaria and Daucus carota, for which more DMPs were found in clusters not having significant blast hits to genes. CHG DMPs are absent in Daphnia magna and are most prevalent in Taraxacum officinale, consistent with the observations made for the absence of CHG methylation in Daphnia magna and the higher presence of CHG methylation in Taraxacum officinale (see Fig. 2b).