Rethinking Alzheimer's Disease Therapeutic Systems Using Molecular Dynamics Simulations to Inform and Propel Drug Design.

Abstract

Alzheimer’s disease (AD) is a progressive neurodegenerative condition affecting the brain and central nervous system characterized by the formation of senile plaques and neurofibrillary tangles (NFT), leading to neuronal cell death. Various mechanisms contribute to AD progression, including the accumulated network of hyperphosphorylated tau protein filaments forming NFTs, the aggregation of cleaved amyloid precursor protein (APP) fragments into senile plaques, and oxidative stress-induced disruptions in protein signaling. Central to these processes is the hyperactivity of cyclin-dependent kinase 5 (CDK5), which, under normal conditions, plays a crucial role in neuronal functions such as learning, memory, and cell differentiation when activated by its physiological partners, p35 or p39. However, oxidative stress leads to the cleavage of p35 into P25, resulting in the formation of a dysregulated and hyperactive CDK5-P25 complex. This pathological complex hyperphosphorylates several protein substrates, including tau and APP, leading to the symptoms of neurodegeneration. A promising therapeutic approach involves using a 24-residue peptide, p5, derived from p35, which selectively inhibits the CDK5-P25 complex without disrupting the physiological functions of CDK5-p35. This dissertation uses molecular dynamics (MD) simulations and comprehensive trajectory analysis to explore the various interactions and drug-like properties of p5 at a molecular level to guide the design of p5- derivatives with improved therapeutic efficacy, specificity, and safety. My research investigates multiple binding modes of p5 with CDK5 and P25, indicating both competitive and allosteric inhibition pathways. Additionally, the dissertation addresses any potential off-target effects of p5 on P67 (Munc18-1), a protein involved in regulating neuronal exocytosis, crucial for communication between neurons. Disruption of P67 function can lead to neurodegenerative symptoms similar to those seen in AD and Parkinson’s disease, causing additional treatment and diagnostic challenges. Two groups of binding modes between p5 and P67, exhibiting both allosteric and non-allosteric effects, are identified. The dissertation uses enhanced sampling simulations to further the design of p5 and to explore its selectivity in inhibiting pathological CDK5 hyperactivity without disrupting its physiological function in the presence of P67. Insights for mutagenesis studies and in vitro CDK5 enzyme activity measurements using Bio-FET (Field Effect Transistors) sensors are discussed, establishing a feedback loop between simulations and in vitro experiments for rapid testing of p5-derivatives. Additionally, my research integrates machine learning techniques with protein network representations of MD simulation data to improve the identification and evaluation of reaction coordinates for enhanced sampling simulations in large protein systems.