Model-informed development of RG6102, an amyloid-targeting investigational drug with enhanced brain penetration properties
Joćo A. Abrantes (1), Simon Buatois (1), Carsten Hofmann (1), Niels Janssen (1), Sébastien Jolivet (1), Luka Kulic (1), Hanno Svoboda (1), Nicolas Frey (1), Hanna Silber Baumann (1), Hans Peter Grimm (1)
(1) Roche Pharma Research and Early Development, Pharmaceutical Sciences, Roche Innovation Center, Basel, Switzerland
Objectives: RG6102 is a bispecific 2+1 monoclonal antibody under development for the treatment of Alzheimer’s Disease (AD). It combines the anti-amyloid beta (Aβ) antibody gantenerumab with a “Brain Shuttle” module that binds to the transferrin receptor 1 (TfR1), facilitating the transport across the blood-brain barrier. This work aimed at combining the preclinical to clinical forward translation of pharmacokinetics (PK) of RG6102 with the backward translation of the clinical PK/pharmacodynamics (PD) of gantenerumab to build a model-informed drug development (MIDD) framework.
Methods: The PK time-course of RG6102 and gantenerumab in plasma, cerebrospinal fluid (CSF) and brain following a single intravenous (i.v.) dose was characterized using a model-based approach in non-human primates (NHP; N=30). Correcting brain tissue concentrations for contributions from residual blood was a critical step in the analysis, achieved by the use of a vascular marker. The resulting PK model was translated to humans using allometry and connected to a clinical population PKPD model of Aβ removal, originally built for gantenerumab . The first-in-human (FIH) dose selection was based on a comparison of projected brain exposure of RG6102 with gantenerumab. Thereafter, the PK model was updated using plasma and CSF PK data from the FIH study in healthy volunteers (i.v. doses 0.1-7.2 mg/kg; N=26). Finally, model projections of Aβ removal were used to inform the design of the multiple ascending dose study in patients. Non-linear mixed effects modelling with MONOLIX (NHP data) or NONMEM (clinical data) was used. Advice from the FDA MIDD pilot meeting program was integrated into the modeling framework.
Results: A 2-compartment model captured the plasma PK of RG6102 in NHP. Distribution into CSF and various brain regions was captured using linear 1-compartment models accounting for compound in residual blood in brain tissues in a statistically coherent approach. A higher distribution of RG6102 to brain tissues compared to gantenerumab was found, however, distribution to CSF was similar between compounds. In humans, a 2-compartment model featuring a target-mediated disposition component reflecting binding to TfR1 was used to describe the plasma PK data. The estimated plasma clearance was 2.0 L/day, which was ~1.8-fold faster than the NHP projection. The estimated CSF/plasma AUC ratio was 0.6%, which is 2.5-fold higher than the value estimated in NHP, and higher than the ratio of ~0.1% observed for IgGs . It is likely that both the faster plasma clearance and the higher CSF distribution in comparison to typical IgGs are due to the affinity of RG6102 to TfR1, which is higher in humans compared to NHP. Simulations from the Aβ PKPD model indicate that a dosing regimen of 1.8-3.6 mg/kg Q4W may lead to an Aβ standardized uptake value ratio reduction of ~10-20% after 24 weeks.
Conclusions: A MIDD framework has guided the early clinical development of RG6102 by combining information across compounds and species. In particular, this approach was used to verify the conceptual feasibility of RG6102, guide the FIH doses, and inform the design of the first study in AD patients.
 Frey, N et al. PAGE 25 (2016) Abstr 5795 [www.page-meeting.org/?abstract=5795]
 Freskgård, PO and Urich, E. Neuropharmacology. 2017 Jul 1;120:38-55.