Fig. 3.

Relaxation of polycomb-regulated gene expression. a Relationship between affected transcription units (q < 0.05, cuffdiff) in Fig. 2 and PRC1 or PRC2-related curated gene sets (GSEA MSigDB²²), with significance expressed as q value. Set identifiers and number of genes in sets are indicated. Overlap with genes affected by culture adaption of WT cells (green) and CIZ1-null cells (blue). Overlap with CIZ1-dependent genes in primary cells (dark grey) and culture-adapted cells (light grey). b Schematic of transgenes used to create doxycycline-inducible expression of full-length GFP-CIZ1 in CIZ1 null mice and derived cells⁵. c Heat map showing 266 transcription units (q < 0.05, cuffdiff) affected by the loss of CIZ1 in primary cells, organised by fold-change. Right column, mean fold change in the same transcription units in CIZ1 null cells (13.17 p3, 14.19p1), 24 h after induction of full-length CIZ1⁵ (Supplementary data 4, sheet 2). d Relationship between CIZ1-dependent transcription units in primary cells (q < 0.05, cuffdiff) and all 189 oncogenic signatures (GSEA MSigDB²²), showing overlap (% of genes that are affected by the loss of CIZ1) against significance. Grey dots indicate oncogenic signatures not related to PRC 1 or 2. Red, sets that are up-regulated when a PRC subunit is down-regulated, or vice versa. Yellow, sets that are upregulated when a PRC subunit is upregulated or vice versa. e Heat map showing fold-change in 266 CIZ1-dependent transcription units (q < 0.05, cuffdiff) in primary cells (left, same as Fig. 1c), compared to fold-change in the same genes in culture-adapted derivatives, upon loss of CIZ1 (right). Supplementary data 4, sheet 3. f Heat map showing the 35 transcription units that are CIZ1-dependent in adapted cells (q < 0.05, cuffdiff) (left), and fold change in these transcription units in primary cells (right). Supplementary data 5. g Venn diagram showing overlap between transcription units that are differentially expressed upon loss of CIZ1 in primary and derived lines. Highlighted in Supplementary data 5. All q values (false detection rate corrected P values) for overlap with GSEA MSigDB were calculated using one-sided Fisher’s Exact tests with Benjamini–Hochberg false discovery rate correction. Heat maps are organised by fold-change from up (red) to down (blue)