We have previously obtained compelling proof-of-principle evidence for COX2 gene therapy for fracture repair using integrating retroviral vectors. For this therapy to be suitable for patient uses, a suitable vector with high safety profile must be used. Accordingly, this study sought to evaluate the feasibility of AAV as the vector for this COX2 gene therapy, because AAV raises less safety issues than the retroviral vectors used previously. However, an appropriate AAV serotype is required to provide early increase in and adequate level of COX2 expression that is needed for fracture repair. Herein, we reported that AAV-DJ, an artificial AAV pseudoserotype, is highly effective in delivering COX2 gene to fracture sites in a mouse femoral fracture model. Compared with AAV-2, the use of AAV-DJ led to~ 5-fold increase in infectivity in mesenchymal stem cells (MSCs) and provided an earlier and significantly higher level of transgene expression at the fracture site. Injection of this vector at a dose of 7.5× 10 11 genomic copies led to high COX2 level at the fracture site on day 3 after injections and significantly promoted fracture union at 21 days, as analyzed by radiography and μ-CT. The therapeutic effect appears to involve enhanced osteoblastic differentiation of MSCs and remodeling of callus tissues to laminar bone. This interpretation is supported by the enhanced expression of several key genes participating in the fracture repair process. In conclusion, AAV-DJ is a promising serotype for the AAV-based COX2 gene therapy of fracture repair in humans.