Supplementary Figure 2: Effect of lag on single-cell methylation levels and nascent strand methylation dynamics.

a) Individual cells are binned into M-G₁, G₁-S, S and G₂-M based on expression of known cell cycle regulated genes (Kowalczyk, M.S. et al. Genome Res 25, 1860–72, 2015). Of the 30,311 genes that were ranked according to their differential regulation throughout G₁-S and S-phase, only 1,085 show notable dynamics (Supplementary Data Set 2; Methods, FDR < 0.05). However, no difference in expression for the key DNA methylation regulators is found in ESCs. Expression values displayed were averaged using a moving window of 20 cells. (b) All single cells (n = 455) ordered by mean methylation (based on single cell RRBS) and colored by sorted cell cycle phase. Most cells in G₁ (blue) or G₂-M (green) show higher mean methylation than cells in S-phase (grey). (c) Using single cell RRBS data, CpGs were binned according to whether they were replicated in early or late S-phase (x-axis). Mean methylation for each CpG is displayed as a heat map for cells sorted into early or late S-phase (y-axis), and shows that cells in early S-phase (bottom row) have reduced methylation for early-replicating CpGs, while late-replicating CpGs are still highly methylated. The reverse is observed for cells in late S-phase (top row). Together, our single cell analysis shows that post-replication methylation delay reduces mean methylation of individual cells in S-phase. (d) Heat map of individual CpGs after 1 hour (h) BrdU pulse with varying lengths of chase time (displayed beneath each sample) which shows that methylation levels gradually increase over time. (e) Violin plots display methylation levels for CpGs that are replicated in early (S1 + S2), mid (S3 + S4) and late (S5 + S6) S-phase; see Fig. 1b. For each subset of CpGs, the methylation levels for 0, 1, 4 and 16 h are shown. (f) Bars display all matched CpGs between replicates across the time course. The color indicates the absolute difference in methylation between matched CpGs. In contrast to the variable correlation between individual CpGs, the global mean methylation differences between replicates are very low (0.04, <0.01 and <0.01 between the 1, 4 and 16 h replicates, respectively; shown in Fig. 1e). (g) Relationship between methylation values of neighboring CpGs that were captured in phase on the same sequencing read. The frequency of “transitions” in methylation states between neighboring CpGs is calculated and a transition score is computed by dividing the number of transitions by the number of CpGs. Reads were categorized as “low” if all CpGs showed the same methylation state (number of transitions = 0, score = 0), “medium” for few transitions (score ≤ 0.4), or “high” for frequent transitions (score > 0.4) according to the schematic shown. Examples of reads with their respective score are shown with filled and empty circles representing methylated and unmethylated CpGs respectively. (h) Using the metric described in g, bar plots show the proportion of sequencing reads categorized as low (dark grey), medium (dark blue) and high (light blue) across the time course. Only reads containing ≥3 CpGs were used in this analysis. Several previous studies have reported that DNMT1 is a highly processive enzyme that remains engaged with the same DNA molecule throughout consecutive methylation events (Jeltsch, A. & Jurkowska, R. Z. Trends Biochem Sci 39, 310–318, 2014, Jeltsch, A. & Jurkowska, R. Z. Nucleic Acids Res 44, 8556–8575, 2016). The decrease in the number of transitions between methylation states in our genome-wide read level analysis therefore provides further support for this type of activity during the post-replication window. (i) Correlation between neighboring CpGs on individual sequencing reads according to distance between them is shown for bulk and nascent DNA. The size of the data point is proportional to the number of CpGs.