a Inverse correlation between expression of CCND1 and CCND2 across all MM and normal PC samples. The best linear regression fit is indicated as a blue line with standard error in grey. b Loadings from MOFA analysis for 5 regions and 6 genes within 1 Mb of CCND2 that were selected for MOFA analysis for latent factors 1–5. Both regions and genes ordered in a 5’->3’ direction. c Average chromatin accessibility of regions upstream of CCND2 and CCND2 gene expression across all samples in this study. d Genomic tracks visualization of the CCND2 region. From top to bottom: Hi-C signal for interactions with CCND2 promoter in GM12878 B cells; target location of sgRNAs designed for the CRISPRi experiment; Predicted footprints for MAF TF in MAF-translocated (orange) and MMSET-translocated (green) samples; ChIP-seq signal against MAF, H3K27ac and MED1 in MM.1S cells; normalised ATAC-signal in each subtype, ordered by mean CCND2 expression. e CCND2 expression as assessed by RNA-seq in samples across different MM subgroups and normal donor PC. Subgroups as indicted in adjacent tracks in (d). Boxplots display all values as points, whisker’s box (min to max) and mean expression per subgroup. f CCND2 expression as assessed by qPCR 4 days after CRISPRi of CCND2 super-enhancer and promoter regions. Two sgRNAs were employed to target the promoter (P) and each of four major peaks (1–4) of the CCND2 super-enhancer, as shown in (d), and compared with a non-targeting control (Gal4). Error bars represent mean + SEM for three independent experiments. Statistical analysis: one-way ANOVA with Dunnett’s post-hoc multiple comparisons correction. *p < 0.05, ***p < 0.001, ****p < 0.0001. g Heatmap illustration of TF motifs significantly enriched in CCDN2high versus CCND2low HD samples, as identified by differential footprinting analysis. Differential footprints were identified in peaks 3 and 4 (see Fig. 5d).