Dynamic database behaviour can be specified by dynamic integrity constraints, which determine admissible sequences of database states, and by transaction specifications, which induce executable sequences. Constraints are expressed by formulae of temporal logic, whereas transactions are defined by pre⧸postconditions in predicate logic. This article presents concepts and rules for transforming dynamic constraints into transaction specifications in order to prepare integrity monitoring by transactions. At first, such transition graphs must be constructed from temporal formulae that have paths corresponding to admissible sequences. Then these graphs are utilized to refine and to simplify pre⧸postconditions systematically so that every executable state sequence will become admissible, too.