(M1130-10-59) Systems Biology Modeling of the Insulin Signaling Network in the Blood-Brain Barrier Endothelium Predicts Cerebrovascular Inflammation in Alzheimer’s disease

Purpose: Epidemiological evidence has suggested a strong association between Alzheimer’s disease (AD) and insulin resistance, which is due to disruption in the intracellular insulin signaling network^1,2. Our recent study, using time-series transcriptomics analysis, demonstrated that endothelial insulin signaling regulates critical blood-brain barrier (BBB) functions, including tight junction properties, intracellular trafficking, and immunomodulation³. Therefore, endothelial insulin resistance could lead to BBB dysfunction, which aggravates AD progression. However, insulin signaling is intrinsically dynamic and encompasses interactions among kinetic processes triggered by thousands of proteins. Hence, a mechanistic systems biology model is needed to understand how insulin regulates BBB physiology and its impairment in metabolic syndrome and AD.

Methods: An ordinary differential equation (ODE)-based model was developed composing of two interactive subsystems: 1) insulin signaling transduction in BBB endothelial cells; 2) turnover of vascular cell adhesion molecule 1 (VCAM1), which is a marker for cerebrovascular inflammation. Time- and insulin dose-dependent expression and phosphorylation of various proteins in this network were measured in human cerebral microvascular endothelial (hCMEC/D3) cells using western blots. Kinetic parameters were estimated by fitting the model to both published and newly generated data using Matlab Simbiology^®. The model was further validated using independent datasets obtained from cells overexpressing insulin receptors. The validated model was endowed with proteomics data to simulate insulin signaling and VCAM1 expression under disease conditions. Specifically, a reverse-phase protein array (RPPA) assay was performed to characterize the proteomics changes in the hCMEC/D3 cells exposed to amyloid beta (Aꞵ) peptides. The initial values of various proteins in the signaling model were adjusted based on the RPPA results for model simulation. The work flow of modeling analysis is summarized in Figure 1.

Results: Phosphorylation of insulin receptor (p-IR), insulin receptor substrate-1 (p-IRS1), protein kinase B (p-PKB/AKT) (Figure 2B), endothelial nitric oxide synthase (p-eNOS) was triggered by insulin stimulation in a time- and dose-dependent manner, but the activation of these targets exhibited distinct dynamic patterns. In contrast, VCAM1 expression of hCMEC/D3 cells was reduced by insulin treatment and the effect was found to be mediated via PI3K/AKT/eNOS pathway. Combining insulin signaling cascade and VCAM1 turnover, a mechanistic model was constructed using a system of ODEs (Figure 2A). The model adequately described the experimental results (Figure 2C) and kinetic parameters were successfully estimated with relatively low standard error. Further, the model prediction was verified using independent experimental data of p-AKT and p-eNOS time profiles obtained following insulin stimulation in IR-overexpressed cells. RPPA assay indicated that Aꞵ40 and Aꞵ42 decreased expression levels of various proteins in the insulin signaling network, including IR, IRS1, PI3K, AKT and GSK3 (Figure 3A, B). Model simulation suggested that insulin-stimulated phosphorylation of AKT (Figure 3C, D) and eNOS was significantly dampened in the presence of Aꞵ40/42. On the contrary, VCAM1 expression and associated cerebrovascular inflammation were predicted to be elevated under Aꞵ exposure.

Conclusion: In this study, we successfully developed a systems biology model which quantitatively described the insulin signaling cascade and downstream VCAM1 expression in BBB endothelial cells. Further, the model was used to predict signaling perturbations under Aꞵ exposure using proteomics changes. A similar approach could be applied to transcriptomics changes observed in AD patients to shed light on molecular mediators driving BBB dysfunction in AD.

References: 1 Benedict, C. & Grillo, C. A. Insulin Resistance as a Therapeutic Target in the Treatment of Alzheimer's Disease: A State-of-the-Art Review. Frontiers in neuroscience 12, 215, doi:10.3389/fnins.2018.00215 (2018).
2 Berlanga-Acosta, J. et al. Insulin Resistance at the Crossroad of Alzheimer Disease Pathology: A Review. Frontiers in endocrinology 11, 560375, doi:10.3389/fendo.2020.560375 (2020).
3 Wang, Z., et al. Mapping the dynamics of insulin-responsive pathways in the blood-brain barrier endothelium using time-series transcriptomics data. npj System Biology and Applications (2022). in press

Acknowledgements:This work was supported by National Institutes of Health/National Institute of Neurological Disorders and Stroke R01NS125437


Figure 1. Work flow of systems biology modeling of insulin signaling network in BBB endothelial cells.


Figure 2. (A) Model structure of insulin signaling cascade and downstream VCAM1 expression. (B) Representative immunoblot of pAKT/AKT at different time points after insulin stimulation in hCMEC/D3 cells. (C) Model fitting of pAKT/AKT-time profile.


Figure 3. (A-B) Log2 fold change of protein expression upon (A) Aꞵ40 and (B) Aꞵ42 exposure in hCMEC/D3 cells. (C-D) Model simulation of dose dependent pAKT profile in (C) Aꞵ40 and (D) Aꞵ42 treated hCMEC/D3 cells.