Discussion
Due to potential long-term safety risks, FIH study of most GT products is a dose-finding study conducted in a dozen or so patients with a rare disease of genetic disorders. Each patient usually receives a single dose of AAV vector only because the immune responses caused by repeated dosing may render subsequent administrations ineffective9. A patient receiving a low and ineffective starting dose may be ineligible to receive an escalated dose due to activated immunity against this AAV vector. Thus, an accurate prediction of FIH dose is critical for clinical development of GT products. In addition to the direct vg/kg conversion approach widely used in clinical development of AAV GT16 and recently reported allometric scaling approach,5 this study demonstrated that interspecies dose-response normalization could be used as an additional approach for FIH dose prediction for AAV-mediated GT. This new approach bypasses pharmacokinetic data and extrapolates dose-response relationship from preclinical species to hemophilia patients. This approach is especially useful at the early stage of clinical development when clinical data are insufficient to support the development and validation of complex quantitative systems pharmacology models. However, the performance of the interspecies dose-response normalization approach can be compromised by poor quality of preclinical data, uncertainties in human immune responses to vector/transgene product, and substantial interspecies differences in vector transduction efficiency and transgene expression efficiency.
The quality of preclinical data is critical for successfully extrapolating dose response from animals to humans. Preclinical dose-finding studies for AAV-mediated GT products are usually conducted in dogs or monkeys with a small sample size (N = 2 – 6 per dose level) and within a limited dose range. Due to the large inter-individual variability in transgene product levels, a preclinical dose-finding study with appropriate sample size is recommended. As shown in our regression analysis, the total vector dose range often did not overlap between dogs/monkeys and humans, which might compromise the quality of interspecies dose-response extrapolation. This issue can be addressed by (1) including large animal models with a body weight comparable to human (i.e., swine and cattle) into does-finding studies, and (2) testing GT products in the large animal models with a wide dose range. Since intracellular synthesis rate of transgene product is reversely correlated with W0.25, large animal models are anticipated to be more representative of humans in term of transgene product expression efficiency. The quality of both allometric scaling and interspecies dose-response normalization are expected to be improved by including dose response data of large animal models.17
Immune responses to vectors, transgene products, and AAV capsid antigens expressed on the surface of transduced cells in animals and humans are a big challenge in extrapolating dose-response from animals to humans.8 AAV capsid-specific T cell responses were detected in hemophilia B patients receiving moderate- or high-dose rAAV2-hAAT-FIX or scAAV2/8-LP1-hFIXco vectors. The presence of T-cell responses might eliminate vector-transduced human hepatocytes, decrease transgene product expression, and thus lead to substantial underprediction of FIH dose.8 Several steps may be taken to reduce the risk of AAV capsid-specific T cell responses: (1) develop appropriate in vitro tests (i.e., enzyme-linked immunospot assay and cytotoxic T lymphocyte assay) to identify and eliminate vector candidates with a high risk of T cell responses16; (2) exclude patients with preexisting immunity from FIH studies; and (3) maximize vector potency through capsid engineering and use high-activity transgene product variant and thus reduce FIH dose because AAV capsid-specific T cell responses are dose-dependent.8, 18
With the advance in capsid engineering and availability of new potent promoters/enhancers, some bioengineered AAV vectors showed a much higher transduction efficiency and a similar or even higher transgene expression efficiency in human hepatocytes than that in animal hepatocytes,15 which might cause an overprediction of FIH dose using allometric scaling or interspecies dose-response normalization approaches. The large interspecies differences in transduction efficiency and transgene expression efficiency can be detected using a humanized mouse liver model.15 These interspecies differences should be considered when dose response is extrapolated from animals to humans.
In this study, the proposed interspecies dose-response normalization was assessed with AAV vectors for hemophilia therapy only. Further work is needed to assess if this approach can be applied to (1) AAV-encoding transgene products that function in systemic circulation for other diseases; (2) other intravenously administered viral (e.g., adenoviral and lentiviral vectors) and non-viral GT modalities (e.g., lipid nanoparticles); and (3) locally administered GT products (e.g., intrathecal or intravitreal administration).