The ideal gene therapy vector should enable persistent expression without the limitations of safety and reproducibility. We previously reported that a prototype plasmid vector, containing a scaffold matrix attachment region (S/MAR) domain and the luciferase reporter gene, showed transgene expression for at least 6 months following a single administration to MF1 mice. Following partial hepatectomy of the animals, however, we found no detectable vector replication and subsequent propagation in vivo. To overcome this drawback, we have now developed an in vivo liver selection strategy by which liver cells transfected with an S/MAR plasmid are provided with a survival advantage over non-transfected cells. This allows an enrichment of vectors that are capable of replicating and establishing themselves as extra-chromosomal entities in the liver. Accordingly, a novel S/MAR plasmid encoding the Bcl-2 gene was constructed; Bcl-2 expression confers resistance against apoptosis-mediated challenges by the Fas-activating antibody Jo2. Following hydrodynamic delivery to the livers of mice and frequent Jo2 administrations, we demonstrate that this Bcl-luciferase S/MAR plasmid is indeed capable of providing sustained luciferase reporter gene expression for over 3 months and that this plasmid replicates as an episomal entity in vivo. These results provide proof-of-principle that S/MAR vectors are capable of preventing transgene silencing, are resistant to integration and are able to confer mitotic stability in vivo when provided with a selective advantage.