Development of a semi-mechanistic pharmacokinetic/pharmacodynamic model of a small interfering RNA targeting liver proteins in mice and cynomolgus monkeys
Jennifer Lang (1a), Jan-Stefan Van der Walt (2), Thomas Beyer (3b), Gregory Lackner (3c), Christopher Wiethoff (3d), Parag Garhyan (3a) and Aurélie Lombard (4a)
(1) Eli Lilly & Company, Bracknell, UK; (2) Occams, Amstelveen, The Netherlands; (3) Eli Lilly & Company, Indianapolis, US; (4) Eli Lilly & Company, Neuilly, France; (a) Global PK/PD and Pharmacometrics; (b) Biology, Dyslipidemia & Outcomes Research; (c) RNA Therapeutics and Discovery; (d) Biologics Investigational ADME.
Introduction: Small interfering ribonucleic acid (siRNA) is an emerging class of targeted therapies that aims to regulate the expression of specific genes by preventing the protein translation process. LY is a tris N-acetylgalactosamine (GalNAc)3 conjugated siRNA designed to target a specific liver protein involved in triglyceride metabolism. Following subcutaneous administration, GalNAC-siRNAs are rapidly internalised by hepatocytes via endocytosis and released in the cytoplasm as free siRNAs. Then, free siRNAs bind to the RNA-induced silencing complex (RISC), leading to the target mRNA degradation. The interspecies translation of PK and PD is challenging due to the differences between the short systemic exposure (i.e., plasma half-life ~2-10 hours) and the prolonged pharmacodynamic activity (i.e., several weeks to months) in addition to the lack of mechanistic understanding. The objectives of this study were (i) to develop a semi-mechanistic PK/PD model using mice data and (ii) to translate and optimise the mouse model to cynomolgus monkeys using allometric scaling.
Methods: A semi-mechanistic PK/PD model was developed in a stepwise approach in NONMEM v7.5.0 using FOCE-I. Plasma, total liver and RISC-bound concentrations of LY, liver target mRNA and serum protein levels measured in mice following a single SC dose of LY at 3 different dose levels were used for model development [1]. A one-compartment model was developed to describe the LY PK concentrations in the systemic circulation. Slow liver uptake and endocytosis were considered in the liver compartment to account for the prolonged tissue residence time of total liver concentrations. Then, a fraction of free siRNA was estimated to bind to the RISC. A direct response model using RISC-bound concentrations was then implemented for the inhibition of target mRNA. Finally, target mRNA in the liver was linked to triglyceride degradation in the systemic circulation.
Subsequently, the semi-mechanistic model developed in mice was used for interspecies translation from mouse to cynomolgus monkey. Standard allometric scaling based on body weight was first considered on clearance and volume. Parameter estimation was also applied to improve the description of cynomolgus monkey PK and PD data (i.e., plasma, total liver and RISC-bound concentrations of LY, target mRNA in the liver and serum protein) after a single SC dose of LY at one dose level. To this end, optimisation of allometric scaling factors, physiological processes (e.g., turnover rate, uptake clearance) and response model parameters was explored.
Results: The semi-mechanistic PK/PD satisfactorily described PK and PD data in mice at different doses in the systemic circulation and target liver tissue. Nonlinearity in plasma PK was considered using a Michaelis-Menten saturable elimination. Total liver and RISC-bound PK were best described using non-saturable first-order processes. Predicted maximum target mRNA knockdown in mice was dose-dependent, in line with the observations (mean predicted/mean observed = 43.7/35.7% and 81.2/86.5 for low and high doses, respectively).
For the cynomolgus monkey, estimated allometric scaling factors on plasma clearance and volume parameters were comparable to standard allometric exponents (0.692 vs 0.75 and 1.18 vs 1 for clearance and volume, respectively [2]). Overall, the data in cynomolgus monkeys were well predicted by the mouse model and no optimisation was needed for turnover rates. However, slower absorption was observed in cynomolgus monkeys (0.355 vs 0.231 h-1 in mice and cynomolgus monkeys, respectively). Estimated fraction of siRNA that binds to the RISC in cynomolgus monkeys was higher than the estimated value in mice (0.0013 vs 0.0077 in mice and cynomolgus monkeys, respectively). Predicted maximum mRNA KD was consistent with the observed value (mean predicted/mean observed=86.1/84.5%).
Conclusion: The semi-mechanistic PK/PD developed in mice was successfully translated to cynomolgus monkeys. Allometric scaling allowed adequate characterisation of interspecies changes in clearances and volumes of plasma and target tissue. Minimal optimisation was needed to account for differences in absorption and minor changes in physiological processes for this specific target were shown between mice and cynomolgus monkeys. This interspecies semi-mechanistic PK/PD model provides a model-informed drug development (MIDD) framework to support dose selection for first-in-human study.
References:
[1] V.S. Ayyar, D. Song, S. Zheng, T. Carpenter, D.L. Heald, J Pharmacol Exp Ther 379 (2021) 134–146.
[2] Y. Huh, D.E. Smith, M.R. Feng, Xenobiotica 41 (2011) 972–987.
