Tatiana Karelina, PhD: No financial relationships to disclose
Oleg Demin Jr: No financial relationships to disclose
Background. Several published quantitative systems pharmacology (QSP) published models describe the dynamics of amyloid (Aβ) plaque reduction during anti-amyloid immunotherapy. However exact mechanisms linking reduction of amyloid with cognitive effect were not described, and moreover, dynamics of other important biomarker (MRI brain volume) contradicts to hypotheses of correlation between brain atrophy and cognitive score (MRI volume demonstrates reduction vs placebo, which is not in line with cognitive improvement). Our goal was to develop model of key players of AD pathology (Aβ, tau pathology and microglia (MG) activation) and key neurotoxicity mechanisms allowing for description of widely used cognition scores (CS) during anti-amyloid treatment.
Methods. Model includes Aβ production, distribution, and aggregation [1]. Tau protein pathology was described semi-empirically as hyper-phosphorylation probability. Functionality of neurons is associated with the content of synaptic proteins (variable SP in the model), which determines neuronal plasticity (long-term potentiation (LTP) in electrophysiological experiments). In the model, SP is reduced by Aβ, phospho-tau, Il1β (secreted by MG) through the action of Calpain and p38 kinase. Clinical CS, e.g. MMSE etc., are described by empirical functions of atrophy (reduction of number of live neurons) and reduction of SP. The model was calibrated on the baseline levels of Aβ and tau in brain and CSF, data on MG phenotypes, electrophysiological data in transgenic mouse AD models in presence of different treatments, and on MRI brain atrophy data. Then CS functions were calibrated and validated on aducanumab clinical data [2]. Uncertainty of parameters for neuronal SP function (effect of Aβ and Il1β on calpain and p38) was determined on mouse data, then, for AD patient, variabilities for several model components were added, based on collected data.
Results. Model captures difference between AD and control subjects for Aβ and tau biomarkers and MG activation. It describes LTP change in Aβ and tau Tg mice and partial restoring of LTP by treatments (calpain inhibitors, anti-Il1β). This calibration resulted in successful translation and validation on relative reduction of neuronal proteins in CSF in AD subjects [3] and correlation between MEP amplitude and CSF biomarkers (Aβ and tau) in AD subjects. After description of Aβ reduction by aducanumab, model reproduces reduction of tau in CSF and brain volume dynamics. After calibration of CS dynamics on placebo data, model correctly reproduced MMSE and CDR, although additional calibration was required for ADAS. Model also correctly reproduced verubecestat clinical data.
Conclusions. Modeling of key neurotoxicity mechanisms allows for semi-mechanistic translation from electrophysiological studies in mouse to human clinical trial data. Potentially it can be used for prediction of clinical trial results for other targets alternative to amyloid.
