A proximity-based in silico approach to identify redox-labile disulfide bonds: the example of FVIII

Background: Disulfide bonds (DSBs) are increasingly recognised as important regulators of protein activity and can influence a broad range of pathophysiological processes. Coagulation factor VIII (FVIII) is an essential clotting factor whose lack of function results in haemophilia A, a rare bleeding disorder that manifests with prolonged bleeding time, bruising, and joint bleeds. FVIII activity is corrected by administration of exogenous FVIII. It has been known for ~40 years that FVIII activity is modulated by its redox state, presumably through the reduction of labile DSBs. Identifying redox-labile DSBs can provide valuable information about the behaviour of a molecule under certain pathophysiological conditions as well as help guide the development, manufacture, and storage of biological medicines. Aims: We hypothesised that in silico modelling could predict the redox lability of DSBs in FVIII and identify those susceptible to reduction. Methods: Applying all-atom parallel bias metadynamics simulations and a proximity criterion, we developed a model using experimentally validated labile DSBs. We then applied this to the FVIII light chain using TCEP as the redox catalyst to establish a hierarchy of redox labile DSBs, which we then evaluated using quantitative mass spectroscopy (qMS). Results: We found that redox-labile disulfides were identified as those to which the most TCEP molecules could approach and get within 0.8nm. This proved accurate for DSBs in TF (186-209), CD44 (77-97), CD132 (160-209), and IL-4 (3-127). Having already experimentally validated that FVIII was redox-labile and that the light chain (A3-C1-C2) was reduced by TCEP and thioredoxin, we used our model to predict which DSB was likely reduced. The model predicted DSB 1899-1903 to be the most redox labile, which we subsequently confirmed by qMS. Conclusions: We can accurately predict redox-labile disulfide bonds in silico by using a proximity parameter.