Nitrogen mustards play an important role in current cancer chemotherapy. The most effective antitumour agents are those carrying two alkylating functions, probably through their ability to form interstrand cross-links in DNA. Such lesions appear to create more of a block in DNA replication and are more difficult to repair than are most monoadducts. Although there were early reports that monofunctional drugs were more mutagenic than the bifunctional drugs, this has not been formally proved using structurally related drugs in a mutagenicity assay capable of detecting a range of different events. We have studied both the mutagenic potency and spectrum of events caused by treatment with the clinical agent, chlorambucil, compared with its half-mustard analogue, in Chinese hamster ovary (CHO)-AS52 cells. Although both drugs caused comparable increases in mutation frequency at doses killing 90% of cells (from around 9x10-6 to around 9x10-5 mutant cells), the nature of events differed significantly between the drugs. By far the majority of mutations caused by the half-mustard were transversion mutations, and almost all of these could be interpreted in relation to the DNA adducts that are known to be formed. In contrast, the majority of chlorambucil-induced mutations were major deletions, and point mutations were only identified from a few clones. Parallel micronucleus assays verified that chlorambucil has a stronger ability to break chromosomes than the half-mustard. These two drugs are thought to form similar monoadducts, but only the full mustard can form interstrand cross-links. The data suggest that DNA cross-links, although only a minor fraction of the total lesions, dominate the mutagenic spectrum and lead to gross changes at the chromosome level that can not be readily associated with individual lesions produced by the drug.