Development of an Agent-Based Model to Investigate Response Durability in Hemophilia B Patients Treated with the AAV-Based Coagulation Factor IX (FIX) Gene Therapy Etranacogene Dezaparvovec

Introduction

Adeno-associated virus (AAV) based gene therapy offers a durable, single treatment solution for enzyme deficiencies that affect the hematologic system. AAVs deliver DNA sequences into target cells, where the viral genome forms stable episomes, enabling the expression of therapeutic protein(s) by the host cell's transcriptional and translational machinery. For self-renewing (i.e., not terminally differentiated) tissues or organs, the effects of cell turnover and episome propagation on durability of the therapeutic effect remain uncertain.

Differential equation-based models have previously been developed to predict AAV uptake and protein production. However, episome loss associated with target cell turnover has proved difficult to capture in these models. To overcome this shortcoming, an agent-based model (ABM) was developed to predict the long-term durability of coagulation factor expression for gene therapy for hemophilia and described herein. ABM enabled simulation-based analyses on the impact of assumptions around cell biology, patient physiology and product characteristics on the durability of gene therapy.

Methods

This ABM was applied to etranacogene dezaparvovec (Hemgenix), a liver-targeting gene therapy for hemophilia B, that is based on an AAV5 vector delivering a transgene encoding the Padua variant (R338L) of human FIX. In this model, each agent represents an individual hepatocyte capable of cell division and death, and of being transduced with one or multiple AAV particles. Transduced target cells may pass episomes from mother to daughter cells during simulated cell division while all episomes were assumed lost upon cell death. The resultant model is specified by nine kinetic parameters describing aspects of liver cell biology and AAV-mediated gene delivery: cell proliferation and death, diploid to polyploid differentiation, diploid vs. polyploid composition as a function of patient age, AAV-transduction efficiency, effect of active episomes, and associated FIX transcription, translation and degradation rates.

Kinetic parameters were informed by literature estimates, and the model was used to simulate FIX activity over 20 years following treatment. Population variance in key biological parameters was accounted for by random uniform sampling: initial fraction of infected cells (0.1-0.7), number of episomes per transduced cell (10-1000), patient age (20-80 years) and the maximum number of functional episomes expressed per cell (1-5). 160,000 virtual patient simulations using a 1000 agent-based representation of liver tissue resulted in reproducible statistics. Model development, simulations and analyses were conducted using Julia 1.10.4.

Results

ABM simulations of FIX activity in patients treated with etranacogene dezaparvovec were consistent with clinical observations over three years follow-up from the ongoing Phase 3 pivotal HOPE-B study (NCT03569891). Median FIX activity of 28% (95% CI: 6.1-84%) of normal levels was predicted by year three post-treatment with etranacogene dezaparvovec, compared to an observed median value of 36% (range: 4.8-80%, n=48). Prospective simulations predict the etranacogene dezaparvovec-induced FIX activity to gradually stabilize at population median of 23% of normal levels (95% CI: 4.6-71%) over 20 years. Sensitivity analyses identified liver cell loss in general as potentially limiting the duration of FIX expression.

Conclusions

There are many uncertainties around the long-term outcomes of AAV-based gene therapies. This work demonstrates that ABMs can be used to predict how biological variables such as target cell turnover, liver physiology and viability along with product characteristics may impact response durability. While the model is applied herein to the context of AAV gene therapy-based FIX for hemophilia B, the computational framework should be adaptable to many other AAV-based gene therapies and diseases.

Li:Metrum Research Group: Current Employment. Nandy:CSL Behring: Current Employment. Jordie:Metrum Research Group: Current Employment. Peppel:CSL Behring: Current Employment. Knab:Metrum Research Group: Current Employment. Kirouac:Metrum Research Group: Current Employment. Wilkins:Metrum Research Group: Current Employment. Nuthalapati:CSL Behring: Current Employment.