Background Adenoviral vectors are very attractive agents for use in in vivo arterial gene transfer. In a previous study, we demonstrated high-efficiency adenovirus-mediated gene transfer into medial smooth muscle cells of balloon-injured rat carotid arteries. We now further characterize this system by investigating the reproducibility of recombinant gene expression, the presence of acute adenovirus-associated toxicity in the vessel wall, and the optimal virus concentration for transduction. Methods and Results Balloon-injured rat carotid arteries were incubated with (1) an adenovirus expressing a β-galactosidase gene (Av1LacZ4), (2) a related adenovirus without the recombinant gene (Addl312), or (3) control solutions. Recombinant gene expression was determined 3 days after gene transfer by measurement of β-galactosidase activity in vessel extracts and by counting of smooth muscle cells in microscopic sections that were histochemically stained to detect recombinant β-galactosidase activity. Adenovirus-associated toxicity was assessed in microscopic cross sections by counting of total smooth muscle cell nuclei in the media (to identify cell loss) and characterization of medial cellular infiltrates with histochemical stains for specific inflammatory cells (neutrophils, lymphocytes, macrophages, and monocytes). Maximum recombinant gene expression after incubation with Av1LacZ4 was produced by virus concentrations ranging from 2×10 10 to 5×10 10 plaque-forming units (pfu)/mL. Surprisingly, use of a higher concentration of Av1LacZ4 virus (1×10 11 pfu/mL) resulted in loss of recombinant gene expression. In addition, infusion of either Av1LacZ4 or Addl312 at 1×10 11 pfu/mL resulted in statistically significant decreases in medial smooth muscle cell number (53% decrease, P <0.01 for Av1LacZ4; 39% decrease, P <.05 for Addl312) compared with vessels infused with control solution. This decrease in smooth muscle cell number was not present after the infusion of virus at lower concentrations. The number of neutrophils in vessel cross sections was significantly increased (fivefold; P <.05) after infusion of Av1LacZ4 at 1×10 11 pfu/mL compared with vessels infused with control solution. Lymphocytes, macrophages, and monocytes were present only in low numbers in all vessel cross sections and were not increased consequent to adenovirus infusion. Conclusions This model of focal in vivo adenovirus-mediated gene transfer into the media of injured arteries is highly reproducible and allows high-level recombinant gene expression over a fairly narrow range of virus concentrations. Acute adenovirus-associated tissue toxicity, as demonstrated by medial smooth muscle cell loss and neutrophilic infiltrates, places an upper limit on virus concentration and associated recombinant gene expression and suggests the presence of a “window” of virus concentration in which either therapeutic or biological effects of recombinant genes can be studied in the absence of associated acute toxicity.