Introduction
With the approvals of Luxturna® and Zolgensma® in Europe and the United States, recombinant adeno-associated virus (rAAV)-based gene therapies (GT) have shown promise for the treatment for diseases with genetic disorders. Establishing dose response relationships in targeted patients is foundational in the development of therapeutic drugs including GT. The efficacy and safety of adeno-associated virus (AAV)-based GT are largely dose-dependent in both preclinical species and humans.1 However, it is still challenging to extrapolate dose response of GT from animals to humans. For example, dose response curves in preclinical species and humans are usually nonlinear and does not follow a consistent pattern among species.2 Traditional quantitative methods such as PK-PD modeling used to predict human dose-response relationships of small molecules and therapeutic proteins are difficult to apply to GT, and model structures and parameter paradigms may not be directly applicable to these complex GT products.3 A mechanistic modeling approach incorporated with disposition of vector in the body, transduction efficiency in target tissues, expression strength in transduced cells, and duration of expression has been proposed to establish human dose response relationships for GT.2However, our current understanding of GT pharmacology is still limited, and experimental data are usually inadequate to validate a complex mathematical model for GT products especially at early development stage.
It was reported that the metabolic rate of cells (Bc) in vivo in mammalian species decreased with increasing body weight (Bc = B0 / W0.25, where B0 is the cellular metabolic rate expressed as 3 × 10−11 Watts per cell for organism with mass of 1.0 gram, and W is the mass of a mammalian species).4Since intracellular synthesis of transgene protein following GT is dependent on metabolic rate of cells, Tang et. al. proposed a concept of gene efficiency factor (GEF) to describe the efficiency of gene transfer system and demonstrated a linear relationship between logGEF and logW using preclinical and clinical data of three AAV vectors for hemophilia B therapy, where W was the body weight of mammalian species.5 Tang et. al. concluded that body weight-based cross-species allometric scaling of GEF could be used to predict first-in-human (FIH) dose of AAV-mediated hemophilia B GT. However, Aksenov et. al. pointed out the limitation of Tang’s allometric scaling approach that Tang et. al. assumed a linear dose response relationship for serum factor IX (FIX) while the dose-response curves observed in most preclinical species and hemophilia B patients were nonlinear.2 Then, Aksenov et. al. used a powder regression model (FIX concentration = a × Doseb) to describe dose-FIX concentration relationships for AAV-FIX vectors, where a was proportionality coefficient and b was the exponent for dose. Aksenov et. al.’s analysis showed that the dose response curves of AAV vectors did not follow a consistent pattern across species and no obvious relationship for proportionality coefficient or exponent vs. body weight was observed. Therefore, Aksenov et. al. concluded that Tang’s allometric scaling was unable to accurately predict human dose for AAV GT.
Enlightened by interspecies normalization of plasma drug concentration-time curves using Dedrick plot,6 this author hypothesized that interspecies normalization approach might be applied to dose-response extrapolation from animals to humans. The Detrick plot approach, also called species-invariant time method, assumes that by normalizing the concentrations with body weight and transforming chronological time to physiological time, the plasma drug concentration–time curves should be superimposable in all species.7 The transformed concentration–time curves of various species are superimposed and then back-transformed to estimate human plasma concentration–time profile. Since the expression efficiency of a transgene product is reversely correlated with W0.25, for a transgene product that functions in systemic circulation, this author hypothesized that the total amount of transgene product in blood circulation across species could be normalized to a species-invariant scale using an exponent of 0.25. The normalized transgene product–dose curve can be back-transformed to predict human dose-response curve.
Among different types of AAV programs, hemophilia GT has more preclinical and clinical dose-response data available in literature. The most common types of hemophilia are hemophilia A and hemophilia B, caused by mutations in F8 or F9 , coding for factor VIII (FVIII) and factor IX (FIX) proteins, respectively.8In this study, this author explored if the total serum amount of FVIII or FIX following AAV-mediated GT could be normalized across species and if the normalized dose response could be extrapolated to hemophilia patients. Furthermore, the normalized serum FVIII or FIX-dose curves were used to predict the FIH dose of AAV vectors. The predictive performance of this interspecies normalization approach was compared to that of two previously reported approaches, direct vg/kg conversion from preclinical doses and allometric scaling.