Fig. 4
ZFP296 knockout decreases NuRD binding genome-wide. a Volcano plot from label-free whole proteome mass spectrometry analysis of WT and Zfp296 KO clone 27 ESCs, graphed as in Fig. 2a. The CRISPR KO target Zfp296 is colour-coded in green, NuRD core subunits are colour-coded in red. b Heat map showing the ChIP-seq read density for MBD3 and CHD4 in WT and Zfp296 KO clone 27 ESCs, centred on the union of MBD3 peaks from ESC and NPC. c Band plots of MBD3 and CHD4 ChIP-seq in WT and Zfp296 KO clone 27 ESCs at MBD3 binding sites for each class of NuRD binding. d Fold change in ChIP-seq read counts (Zfp296 KO clone 27/WT) of MBD3 and CHD4 at MBD3 binding sites for each class of NuRD binding. Box: median (central line), first and third quartile (box limits); whiskers: 1.5  × interquartile range. ****p < 0.0001, as assessed by ANOVA. e Scatterplot of MBD3 ChIP-seq read counts at ESC-enriched MBD3 binding sites in WT and Zfp296 KO clone 27 ESCs, coloured for the amount of ZFP296 at the same locus. The diagonal line indicates no change between the two conditions. A linear model on these data performs significantly better when taking into account ZFP296 levels as a predictive variable (p < 0.001, as assessed by ANOVA). f Boxplots of average MBD3 and CHD4 ChIP-seq signal in WT and Zfp296 KO clone 27 ESCs at all GFP-ZFP296 peaks. Box: median (central line), first and third quartile (box limits); whiskers: 1.5 × interquartile range. ****p < 0.0001, as assessed by Kruskal–Wallis test. See also Supplementary Fig. 3
