The utility of adenoviral vectors for arterial gene transfer is limited by the brevity of their expression and by inflammatory host responses. As a step toward circumventing these difficulties, we used a rabbit model of in vivo arterial gene transfer to test 3 second-generation vectors: a vector containing a temperature-sensitive mutation in the E2A region, a vector deleted of E2A, and a vector that expresses the immunomodulatory 19-kDa glycoprotein (gp19k) from adenovirus 2. Compared with similar first-generation vectors, the second-generation vectors did not significantly prolong beta-galactosidase transgene expression or decrease inflammation in the artery wall. Although cyclophosphamide ablated the immune and inflammatory responses to adenovirus infusion, it only marginally prolonged transgene expression (94% drop in expression between 3 and 14 days). In experiments performed with "null" adenoviral vectors (no transgene), loss of vector DNA from the arterial wall was also rapid (>99% decrease between 1 hour and 14 days), unrelated to dose, and only marginally blunted by cyclophosphamide. Thus, the early loss of transgene expression after adenoviral arterial gene transfer is due primarily to loss of vector DNA, is not correlated with the presence of local vascular inflammation, and cannot be prevented by use of E2A-defective viruses, expression of gp19k, or cyclophosphamide-mediated immunosuppression. Adenovirus-induced vascular inflammation can be prevented by cyclophosphamide treatment or by lowering the dose of infused virus. However, stabilization of adenovirus-mediated transgene expression in the arterial wall is a more elusive goal and will require novel approaches that prevent the early loss of vector DNA.