GM2 gangliosidoses are a family of lysosomal storage disorders that include both Tay-Sachs and Sandhoff diseases. These disorders result from deficiencies in the lysosomal enzyme β-N-acetylhexosaminidase (HexA). Impairment of HexA leads to accumulation of its substrate, GM2 ganglioside, in cells resulting in cellular dysfunction and death. There is currently no treatment for GM2 gangliosidoses. Patients primarily present with neurological dysfunction and degeneration. Here we developed a central nervous system gene therapy through direct injection that leads to long-term survival in the Sandhoff disease mouse model. We deliver an equal mixture of AAVrh8 vectors that encode for the two subunits (α and β) of HexA into the thalami and lateral ventricle. This strategy has also been shown to be safe and effective in treating the cat model of Sandhoff disease. We tested the feasibility and safety of this therapy in non-human primates, which unexpectedly lead to neurotoxicity in the thalami. We hypothesized that toxicity was due to high overexpression of HexA, which dose reduction of vector could not compensate for. In order to maintain AAV dose, and therefore widespread HexA distribution in the brain, six new vector designs were screened for toxicity in nude mice. The top three vectors that showed reduction of HexA expression with low toxicity were chosen and tested for safety in non-human primates. A final formulation was chosen from the primate screen that showed overexpression of HexA with minimal to no toxicity. Therapeutic efficacy studies were performed in Sandhoff disease mice to define the minimum effective dose.