The enzyme 3-methyladenine DNA glycosylase II (AlkA) is a bacterial repair enzyme that acts preferentially at 3-methyladenine residues in DNA, releasing the damaged base. The resulting baseless sugars are alkali-labile, and under the conditions of the alkaline comet assay (single cell gel electrophoresis) they appear as DNA strand breaks. AlkA is not lesion-specific, but has a low activity even with undamaged bases. We have tested the enzyme at different concentrations to find conditions that maximise detection of alkylated bases with minimal attack on normal, undamaged DNA. AlkA detects damage in the DNA of cells treated with low concentrations of methyl methanesulphonate. We also find low background levels of alkylated bases in normal human lymphocytes.Single cell gel electrophoresis (the comet assay) is widely used for the detection of strand breaks in nuclear DNA. It is particularly appropriate for studying the low background levels of damage present in normal human cells, such as peripheral lymphocytes. The cells are embedded in agarose on a microscope slide and lysed with Triton X-100 and 2.5 M NaCl, which remove cytoplasm and most nuclear proteins, but leave the DNA, in supercoiled form, as nucleoids. After incubation in alkali, the DNA is electrophoresed at high pH; DNA is drawn out to form a 'tail' (hence the name 'comet assay') - but only if breaks are present to relax the supercoiling of the nucleoid DNA. In order to increase its sensitivity and selectivity, we have incorporated into the assay an extra step in which the nucleoid DNA is digested with a lesion-specific endonuclease; the additional breaks revealed with this procedure indicate the presence of the particular lesion. So far, endonuclease III (NTH, specific for oxidised pyrimidines) (Collins et al., 1993), formamidopyrimidine DNA glycosylase (FPG, acting on ring-opened purines and the major purine oxidation produce, 8-oxoguanine) (Dušinská & Collins, 1996) and T4 endonuclease V (recognising UV-induced cyclobutane pyrimidine dimers) (Collins et al., 1997b) have been successfully employed. Amongst other things, we have estimated background levels of DNA oxidation (Collins et al., 1997a), and have found this damage to be elevated in human diseases such as diabetes and ankylosing spondylitis (Dušinská et al., 1999).We now report the use of AlkA, a bacterial repair enzyme whose main substrate is 3-methyladenine in DNA, though it also recognises - with lower efficiency - other modified bases (Lindahl, 1993). A recent report (Berdal et al., 1998) suggests that repair enzymes supposedly specific for alkylated bases may in fact create breaks non-selectively (though much less efficiently) at normal bases. Given the size of the genome, even a low efficiency of non-specific breakage could significantly interfere in estimations of background levels of alkylation damage. We reasoned that, by employing a range of concentrations of the enzyme, and carrying out incubations for different lengths of time, we might find a concentration at which only the alkylated bases would be detected, so that the number of breaks would increase to a certain level and then plateau. After optimising reaction conditions, we tested the assay on lymphocytes from different individuals, and also, as a positive control, examined alkylation damage induced by methyl methanesulphonate.