As a consequence of their mechanism of action, DNA (cytosine-5) methyltransferases from both prokaryotes and eukaryotes necessarily recognize mispaired bases in unusual DNA structures as catalytic transition-state analogs. A review of the available data suggests that the enzymes are designed to stall at these sites because they are unable to release substrates or products that are fixed in a conformation resembling the transition state. The enzymes can operate by a two-step process in which they first methylate extrahelical cytosines satisfying their recognition requirements and subsequently stall at the site of methylation. On RNA and DNA RNA hybrids they may operate by a similar one-step process in which they stall at transition-state analogs without methylating cytosine moieties. These natural capacities suggest that the enzymes may physically participate in stable nucleoprotein assemblies formed as components of normal chromatin structure or as intermediates in the repair of unusual structures. The methyltransferases, themselves, may physically participate in chromosome remodelling as part of a mechanism of inactivation or imprinting by stabilizing RNA DNA hybrids or RNA RNA secondary structure involving cis-acting untranslated RNAs like the product of the Xist gene. Methyl-transferase may physically participate in the repair of certain unusual structures by serving as a nucleation point. The affinity for secondary structure in nucleic acids may account for the spreading of DNA methylation patterns. Titration of host methyltransferase by RNA DNA hybrids and RNA secondary structure formed during retroviral replication in certain tumorigenic retroviruses, like MMTV, may account for global hypomethylation observed in retrovirally transformed cells. In a similar fashion, titration of methyltransferase by secondary structures associated with chromosome instability may account for global hypomethylation observed in association with local hypermethylation in tumorigenesis.