Abstract

In addition to its well-documented effects on gene silencing, cytosine methylation is a prominent cause of mutations. In humans, the mutation rate from 5-methylcytosine (m5C) to thymine (T) is 10-50-fold higher than other transitions and the methylated sequence CpG is consequently under-represented. Over one-third of germline point mutations associated with human genetic disease and many somatic mutations leading to cancer involve loss of CpG. The primary cause of mutability appears to be hydrolytic deamination. Cytosine deamination produces mismatched uracil (U), which can be removed by uracil glycosylase, whereas m5C deamination generates a G x T mispair that cannot be processed by this enzyme. Correction of m5CpG x TpG mismatches may instead be initiated by the thymine DNA glycosylase, TDG. Here we show that MBD4, an unrelated mammalian protein that contains a methyl-CpG binding domain, can also efficiently remove thymine or uracil from a mismatches CpG site in vitro. Furthermore, the methyl-CpG binding domain of MBD4 binds preferentially to m5CpG x TpG mismatches-the primary product of deamination at methyl-CpG. The combined specificities of binding and catalysis indicate that this enzyme may function to minimize mutation at methyl-CpG.