In order to determine the potential of alternative splicing as a means of targeting the expression of therapeutic genes to tumor cells in vivo, a series of episomal plasmid -based ''splice -activated gene expression'' ( pSAGE ) vectors was generated, which contain minigene cassettes composed of various combinations of the three alternatively spliced exons present in the differentially expressed adhesion protein CD44R1 ( v8, v9, and v10 ) with or without their corresponding intronic sequences, positioned in -frame between the CD44 leader sequence and a ''leaderless'' human liver / bone / kidney alkaline phosphatase ( ALP ) cDNA. Because both the v8 -v9 and v9 -v10 introns contain multiple in -frame stop codons, the expression and enzymatic activity of ALP are dependent upon the accurate removal of intronic sequences from the pre -mRNA transcripts encoded by these constructs. The various pSAGE constructs were introduced into CD44H -positive ( T24 ) and CD44R1 -positive ( PC3 ) target cells by electroporation and transfectants selected in hygromycin B. ALP expression was determined by staining with the ALP substrate, BCIP / INT, and the transfected cells tested for their sensitivity to the inactive prodrug, etoposide phosphate. ALP -mediated dephosphorylation of etoposide phosphate generates the potent topoisomerase II inhibitor etoposide. The data obtained indicate that whereas the v8 -v9 intron is spliced in both CD44H -and CD44R1 -positive cells, the v9 -v10 intron is efficiently and accurately removed only in CD44R1 -positive cells. Furthermore, only CD44R1 -positive cells were sensitized to etoposide phosphate when transfected with the v9 -v10.ALP construct. These data emphasize the potential usefulness of alternative splicing as a novel means of targeting gene expression to tumor cells in vivo. Cancer Gene Therapy ( 2002 ) 9, 133 -141 DOI: 10.1038 / sj / cgt / 7700427 Keywords: CD44; alternative splicing; cancer gene therapy G ene therapy shows great promise as a modality in the treatment of cancer.1 As the result of progress made in recent years, there is now no shortage of candidate genes which, if expressed at a high -enough level, can either directly kill tumor cells or sensitize them to the cytotoxic effects of ionizing radiation and/ or chemotherapeutic agents.2 A major outstanding challenge is to develop a safe, effective, and practical means of targeting such genes to tumor cells in vivo while minimizing as much as possible the damage inflicted on normal tissues.3 In this regard, promising results have been obtained using engineered viral vectors that exhibit restricted cellular tropism. 4 Similarly, a number of promoter /enhancer elements have been identified, which can be used to control gene expression in particular cell types or tissues. 5,6 Although little explored, there are additional ways in which gene expression could potentially be targeted to particular cell types in vivo. Specifically, because tumor cells and normal cells differ in their ability to alternatively splice certain pre -mRNA t...