Current base editors (BEs) catalyze only base transitions (C to T and A to G) and cannot produce base transversions. Here we present BEs that cause C-to-A transversions in Escherichia coli and C-to-G transversions in mammalian cells. These glycosylase base editors (GBEs) consist of a Cas9 nickase, a cytidine deaminase and a uracil-DNA glycosylase (Ung). Ung excises the U base created by the deaminase, forming an apurinic/apyrimidinic (AP) site that initiates the DNA repair process. In E. coli, we used activation-induced cytidine deaminase (AID) to construct AID-nCas9-Ung and found that it converts C to A with an average editing specificity of 93.8% ± 4.8% and editing efficiency of 87.2% ± 6.9%. For use in mammalian cells, we replaced AID with rat APOBEC1 (APOBEC-nCas9-Ung). We tested APOBEC-nCas9-Ung at 30 endogenous sites, and we observed C-to-G conversions with a high editing specificity at the sixth position of the protospacer between 29.7% and 92.2% and an editing efficiency between 5.3% and 53.0%. APOBEC-nCas9-Ung supplements the current adenine and cytidine BEs (ABE and CBE, respectively) and could be used to target G/C disease-causing mutations. New base editors change C to A in bacteria and C to G in mammalian cells.