A wide range of cytotoxic and mutagenic DNA bases are removed by different DNA glycosylases, which initiate the base excision repair pathway. DNA glycosylases cleave the N-glycosylic bond between the target base and deoxyribose, thus releasing a free base and leaving an apurinic/apyrimidinic (AP) site. In addition, several DNA glycosylases are bifunctional, since they also display a lyase activity that cleaves the phosphodiester backbone 3' to the AP site generated by the glycosylase activity. Structural data and sequence comparisons have identified common features among many of the DNA glycosylases. Their active sites have a structure that can only bind extrahelical target bases, as observed in the crystal structure of human uracil-DNA glycosylase in a complex with double-stranded DNA. Nucleotide flipping is apparently actively facilitated by the enzyme. With bacteriophage T4 endonuclease V, a pyrimidine-dimer glycosylase, the enzyme gains access to the target base by flipping out an adenine opposite to the dimer. A conserved helix-hairpin-helix motif and an invariant Asp residue are found in the active sites of more than 20 monofunctional and bifunctional DNA glycosylases. In bifunctional DNA glycosylases, the conserved Asp is thought to deprotonate a conserved Lys, forming an amine nucleophile. The nucleophile forms a covalent intermediate (Schiff base) with the deoxyribose anomeric carbon and expels the base. Deoxyribose subsequently undergoes several transformations, resulting in strand cleavage and regeneration of the free enzyme. The catalytic mechanism of monofunctional glycosylases does not involve covalent intermediates. Instead the conserved Asp residue may activate a water molecule which acts as the attacking nucleophile.