Gene therapy vectors based on mammalian promoters offer the potential for increased cell specificity and may be less susceptible than viral promoters to transcription attenuation by host cytokines. The human cytokeratin 18 (K18) gene is naturally expressed in the lung epithelia, a target site for gene therapies to treat certain genetic pediatric lung diseases. Our original vector based on the promoter and 5Ј control elements of K18 offered excellent epithelial cell specificity but relatively low expression levels compared with viral promoters. In the present study, we found that adding a stronger SV40 poly(A) signal boosted primary rat lung epithelial cell expression but greatly reduced cell specificity. Addition of a 3Ј portion of the K18 gene to our vector as a 3Ј untranslated region (UTR) improved epithelial cellspecific expression by reducing expression in lung fibroblasts. The effect of the 3Ј UTR was not related to gross differences in cell-specific splicing. A deletion variant of this UTR further increased lung epithelial cell expression while retaining some cell specificity. These data illustrate the possibilities for using 3Ј UTR to regulate cell-specific transgene expression. Our improved K18 vector should prove useful for pediatric lung gene therapy applications. The epithelial surface of the airways presents an attractive opportunity for the topical delivery of gene-based treatments for pediatric lung diseases. CF and ␣ 1 -antitrypsin deficiency, the two most common fatal monogenic lung disorders in the Caucasian population, have garnered much attention as possible targets for lung gene therapy (1, 2). Other lung disorders that may someday benefit from gene therapies include genetic surfactant protein B deficiency (3) and inflammatory lung diseases (2). CF is caused by a defect in the gene encoding CFTR, a chloride channel residing in the epithelium of the lungs and other organs (4, 5). CF affects multiple organ systems, but it is the lung disease, characterized by thick mucous, persistent bacterial infection, and inflammation, that causes the majority of morbidity and mortality associated with CF (5, 6). In theory, early intervention to deliver a functional CFTR gene to just 5 to 10% of the defective lung cells of CF patients could lead to an improvement in the lung airway surface fluid and an improved prognosis (7,8). Attempts at human gene therapy proved the feasibility of this approach (9), but challenges remain before CF gene therapy becomes a reality.Viral vectors such as those derived from human adenovirus can efficiently deliver DNA to cells, but host immune responses to both the virus and transgene are problematic (10, 11). Complexes of DNA and cationic liposome have been widely tested and have the advantages of relatively low toxicity, low immunogenicity, and ease of preparation (12) but are currently inefficient as DNA delivery vehicles (13,14). It may be possible to counter this problem by choosing a strong expression vector to drive the production of CFTR. It may also be desirable to hav...