Computational analysis of Alzheimer-causing mutations in amyloid precursor protein and presenilin 1

Single-point mutations in the genes coding for amyloid precursor protein (APP) and presenilin 1 (PS1), the active subunit of γ-secretase that cleaves APP to produce Aβ, are the main causes of rare but severe familial Alzheimer's disease (fAD). Recent structures of the transmembrane parts of APP and γ-secretase with a fragment of APP bound enable us to study the origins of the pathogenicity of the single amino acid changes in the context of the actual enzyme-substrate complex, which has not previously been possible. We used the new structures as input for several state-of-the-art computational methods that predict the folding stability effect of mutations. We find that pathogenic mutations almost exclusively reduce the stability of the proteins. Since most "random" mutations of an evolutionarily optimized protein tend to destabilize, we also show that the APP mutations destabilize the complex-bound substrate more than the free substrate, indicating reduced affinity of APP to γ-secretase. We confirmed this using two other methods, BEATMUSIC and mCSM PPI, specifically developed for calculating binding affinities of mutants. Although pathogenic PS1 mutations destabilize the complex and substrate-free form to the same extent, they significantly destabilize the protein more than the control set of random mutations. We conclude that fAD mutations most likely reduce the stability of the protein-substrate complex and thus retention time of APP-C99, leading to premature release of longer toxic Aβ₄₂ in accordance with the FIST model of Aβ production, whereas the observed general destabilization of the protein may reduce activity towards other substrates.