Fig. 3
The role of TET activity in DNA methylation pattern changes. a Principle of TET-mediated differential methylation profiles. When TET-activity is constant, 5mC and 5hmC levels are expected to be correlated (upper panel). If a decrease in TET-activity is causing a CpG to be hypermethylated (lower left panel), a lower than expected 5hmC level is anticipated. If an increase in TET-activity causes CpGs hypomethylation (lower right panel), higher than expected 5hmC levels are anticipated. b Per cell line, hydroxymethylation of all CpGs included on the 450K array binned according to methylation levels, distributed over the x-axis. Bin 1 contains all CpGs with methylation levels between 0% and 10% methylation, bin 2 contains all CpGs with methylation levels between 10% and 20% methylation, etc. All bins containing both variable (hyper-/hypomethylated) CpGs as well as steady CpGs (no change in mC) are depicted. In each bin, averaged hydroxymethylation levels of hyper- or hypomethylated CpGs are depicted as red and blue bar plots respectively and compared to hydroxymethylation levels of CpGs with stable methylation levels, depicted as gray bar plots. Higher 5hmC levels for CpGs in the lower bins (1 to 3) and lower 5hmC levels in the upper bins (8 to 10) for variable CpGs relative to CpGs with steady 5mC levels, indicate that a difference in TET activity may underlie the differential methylation between parental and resistant cells. c For each bin depicted in (b), the difference in average 5hmC level of the stable CpGs minus the average 5hmC level of variable CpGs is plotted on the x-axis. On the y-axis, we depict the corresponding significance of the two-sided Mann-Whitney U test assessing whether the 5hmC levels of stably methylated CpGs differ significantly from the 5hmC levels of variably methylated CpGs (p value, -log10 scale)