Modeling of amyloid accumulation in subjects at risk of Alzheimer’s disease under BACE inhibition treatment.
Etienne Pigeolet (1), Konstantinos Biliouris (2), Ulf Neumann (3), Alexandra Urusova (4), Oleg Demin (4) and Tatiana Karelina (4)
(1) Novartis Pharma AG, Basel, Switzerland; (2) Novartis Institute for Biomedical Research, Cambridge, USA; (3) Novartis Institute for Biomedical Research, Basel, Switzerland; (4) InSysBio, Moscow, Russia
Objectives: The main pathophysiological hypothesis for the development of Alzheimer’s disease currently identifies two abnormal structures suspected to damage neurons in the brain: amyloid beta (Ab) plaques and tangles of tau protein fibers. Several beta amyloid cleaving enzyme (BACE) inhibitors, inhibiting the production of Ab peptides are currently under clinical development.
The pathological time course of biomarkers and clinical symptoms has been characterized [1] indicating that treatments targeting Ab should be administered well before the clinical diagnosis of dementia to have a chance of success.
The slow progression of Alzheimer’s disease prevents rapid assessment of treatment efficacy with the cognitive clinical end-points or amyloid plaque imaging techniques. For treatments aimed at reducing the production of Ab, target engagement is easily measured through Ab CSF concentrations. However, the impact of the Ab reduction on the long term progression of amyloid plaques in the brain is unknown.
The aim of this analysis was to assess, through a systems biology/pharmacology model, what level of BACE inhibition is needed in the long run to stop or slow down the amyloid plaque build-up in the brain.
Methods:
The work was started from an existing model [2]. This system biology model consists of modules describing the synthesis and processing of amyloid precursor protein to Ab species, the distribution of Ab species between biological compartments, their aggregation process and the long term progression of soluble and insoluble Ab. This model was built in a stepwise manner: first from mouse data, scaled up to healthy human subjects with validation using monkey data and then extended to AD patients. It is based on the data and analyses from 73 published references.
Further model development was performed for this research: the mouse model was recalibrated in light of latest literature data on some existing processes and from BACE inhibitors. In house PKPD data from phase I studies with CNP520 and other BACE inhibitors published data were added to calibrate a human model.
Model evaluation was performed through various approaches: identifiability was assessed by log-likelihood profiling, inter-parameter correlations and re-estimations by fixing one parameter at a time. Uncertainty was assessed essentially through model variants exploration. The model was also evaluated through its ability to predict data not used for model calibration.
The model was developed with the DBSolve Optimum software (InSysBio, version 36 ) with fits performed by the DBSolve Maximum Likelihood Estimation method.
Results:
Seven models variants were selected based on different assumptions and goodness of fit. Variations were applied to processes such as synthesis of Ab, destruction of insoluble forms, polymerization, age-dependent changes, presence of Ab42 feedback on Amyloid Precursor Protein production. Two of these variants were further selected for their good fit on the different type of data and for representing model uncertainty. These two models reproduced reasonably well the concentrations of the various amyloid species in plasma and in CSF as well as insoluble forms in brain for healthy subjects and Alzheimer’s patients. They were also reproducing reasonably well the same amyloid species in CSF after BACE inhibitor treatments. Some discrepancies with actual data were however observed: the decrease of Ab42 in CSF between healthy subjects and Alzheimer’s patients for example was estimated at about 90% by the model while the actual data indicate a decrease of about 50%.
To answer to the research question, simulations were performed with 0, 30, 50 and 80% of BACE inhibition for 10 years starting at 65 or 75 years of age. These simulations predicted that the insoluble Ab42 in the brain would reach about 500 to 1100 % or about 65 to 140% of baseline after 10 years of 80% BACE inhibition when started at 65 or 75 years of age respectively. Comparing to a placebo group, differences would be about -70 and -60% respectively.
Conclusions:
The results hint to a monotonic slow down of brain insoluble amyloid with increased level of BACE inhibition. Given the underlying assumptions and uncertainty of the model, validation of this prediction warrants further studies.
References:
[1] Jack, C. et al.(2013) Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers; Lancet Neurol, 12; 207-216.
[2] Karelina et al. (2017) Studying the Progression of Amyloid Pathology and Its Therapy Using Translational Longitudinal Model of Accumulation and Distribution of Amyloid Beta. Clin Pharm Ther P&SP, 6(10);676-685.
