An international collaborative project has been proposed to inactivate all mouse genes in embryonic stem (ES) cells using a combination of random and targeted insertional mutagenesis techniques. Random gene trapping will be the first choice in the initial phase, and gene-targeting experiments will then be carried out to individually knockout the remaining 'difficult-to-trap' genes. One of the most favored techniques of random insertional mutagenesis is promoter trapping, which only disrupts actively transcribed genes. Polyadenylation (poly-A) trapping, on the other hand, can capture a broader spectrum of genes including those not expressed in the target cells, but we noticed that it inevitably selects for the vector integration into the last introns of the trapped genes. Here, we present evidence that this remarkable skewing is caused by the degradation of a selectable-marker mRNA used for poly-A trapping via an mRNA-surveillance mechanism, nonsense-mediated mRNA decay (NMD). We also report the development of a novel poly-A-trap strategy, UPATrap, which suppresses NMD of the selectable-marker mRNA and permits the trapping of transcriptionally silent genes without a bias in the vector-integration site. We believe the UPATrap technology enables a simple and straightforward approach to the unbiased inactivation of all mouse genes in ES cells.