MBD4-deficient cancers exhibit a distinctive mutational signature. (A) Mutation burden in AML, presented as number of base substitutions per exome. Data sourced from dbGaP; cases are ordered on patient identifier (EMC: phs001027¹²  and TCGA: phs000178²⁴ ). (B) Trimer context of C>T mutations in 3 MBD4-deficient AML cases. The center of origin is reflected in the sample label. For comparison, we show signature 1, the established signature associated with 5mC deamination, and all C>T mutations present in TCGA-AML. (C) Schematic representation of MBD4, highlighting germ line loss-of-function variants detected in the AML cases and cases within TCGA (at top). A glycosylase assay was performed to assess the activity of recombinant MBD4 (either AA430-580 or full length), wild-type (WT), delH567, or the catalytically inactive mutant D560A. Substrate (S) and product (P). Consistent results were obtained in 5 experiments for MBD4 AA430-580 and 3 experiments for full length. (D) The proportion of CG>TG mutations observed is set out against the total number of base substitutions detected for all TCGA samples. Samples with germ line MBD4 loss-of-function variants were designated either as heterozygous (monoallelic) or completely inactivated (biallelic) based on the genotype of the cancer (includes somatic mutations). Gray lines mark the top 1% and 0.1% of cases with the highest proportion of CG>TG mutations. A select set of tumor types are highlighted.