It has been observed that overexpression of some short-hairpin RNAs (shRNAs) can induce acute cytotoxicity. This has raised concerns about the safety of using RNA interference (RNAi) technology as a potential therapeutic tool. We have sought to address this challenge of expression control by developing a mono-cistronic vector for the tissue-specific expression of an shRNA from a liver-derived polymerase (pol) II promoter. This new construct efficiently induces target silencing in hepatoma cells in vitro and in mouse livers in vivo. In order to demonstrate the therapeutic potential and improved safety of this approach, we selected an shRNA targeting the envelope surface antigen (sAg) of hepatitis B virus (HBV), which is among the most toxic when expressed from the commonly used U6 promoter. Packaging it as a double-stranded DNA into an adeno-associated virus (AAV) pseudotype 8 and delivering it at a high particle dose (1 x 10(12)) to HBV transgenic mice resulted in the stable reduction of serum sAg to 85% of starting levels, without any concomitant sign of liver damage. With this improved tolerability, the liver-specific pol II shRNA expression persisted for more than one year after the injection. We conclude that this pol II shRNA expression system combined with a potent delivery vector represents an effective alternative to either U6-based strategies or systems that achieve tissue specificity through the use of additional elements.