Adeno-associated viral (AAV) gene transfer to the liver for the treatment of hemophilia B is currently being tested in multiple clinical trials and is also in the pipeline for hemophilia A. Current trials utilize AAV serotype 8, while other capsids including AAV5 will also be tested in hemophilia B patients in the near future. AAV8 shows superior transduction of murine liver and has directed long-term factor IX gene transfer in humans. However, efficacy was substantially lower in human than in murine liver. In order to better predict the performance of the various candidate capsids with liver tropism in humans, "humanized mice" harboring human hepatocytes have been adapted as a pre-clinical model of AAV liver gene transfer. Here, we utilized the liver chimeric FAH model. This model is based on the observation that adult immunodeficient fah-/- mice, which suffer from liver injury when cycled off the tyrosine catabolism blocker NTBC, support the proliferation of human hepatocytes. We used liver chimeric mice highly engrafted with fetal human hepatoblasts (human serum albumin levels 978-3737mcg/ml) or human pediatric hepatocytes (human serum albumin levels 388-3798mcg/ml). We produced AAV vectors expressing enhanced GFP reporter from a self-complementary genome. These included AAV5, AAV8, and also AAV3 (which in recent years has shown high transduction efficiency for human hepatocytes and human liver cancer cell lines in vitro and in murine xenograft models). Vectors were injected into the tail veins of humanized mice at a dose of 1x10^11 vector genomes/mouse. Livers were harvested 14 days later and analyzed for GFP expression in murine and human hepatocytes. Human cells were specifically identified by immunostain for FAH. We analyzed transduction efficiency by immunohistochemical stains of liver sections obtained from formalin fixed tissue (n=3 per vector). Transduced human FAH+GFP+ hepatocytes were visualized by confocal immunofluorescence microscopy followed by analysis with Velocity software. To further substantiate the results obtained by image analysis, we developed a flow cytometry-based method for detection of GFP-transduced FAH+ human hepatocytes and mCD81+ murine hepatocytes. Regardless of the method of analysis, AAV3 was the most efficient in transducing human hepatocytes, transducing 12-15% of human but only 3.5-5.5% of murine hepatocytes. AAV8 consistently transduced 5-6% of human hepatocytes while showing by far the highest efficiency for murine hepatocytes (on average 46%). Since the GFP signal in human hepatocytes transduced with AAV3 was higher than for AAV8 transduced cells, the difference in transduction efficiency of the human cells was likely greater in favor of AAV3 than the approximately 3-fold difference in percent GFP positivity. In contrast, AAV5 performed poorly, transducing only 0.1-0.3% of human hepatocytes and at best 0.8% of murine hepatocytes. This may in part also reflect the previously noted reduction in liver transduction when AAV5 is administered from a peripheral vein. Our model suggests that AAV5 is considerably less efficacious, in particular when given via a peripheral vein. In conclusion, AAV3-based vectors may improve liver-directed human gene therapy for hemophilia compared to the current prevailing AAV8 vector system. DisclosuresHerzog:Novo Nordisk: Research Funding; Spark Therapeutics: Patents & Royalties: Patent licenses. High:Spark Therapeutics, Inc.: Employment, Equity Ownership, Patents & Royalties: AAV gene transfer technology. Srivastava:University of Florida: Patents & Royalties: AAV vector technology.