Modeling biomarker kinetics of Aβ levels in serum following blast

Abstract

Elucidating the unique neuropathological response to blast exposure remains a barrier towards the development of diagnostic approaches for those with blast-induced traumatic brain injury (bTBI). Quantification of biomarker concentrations in the blood post-injury is typically used to inform brain injury severity. However, injury progression and associated changes in biomarker concentrations are sensitive to parameters such as the blast overpressure (BOP) magnitude and frequency of blast exposure. Through this work, a blast-dose biomarker kinetics (BxK) platform was developed and validated for Aβ42 as a promising predictor of injury post-blast. Blast-dose responses accounting for BOP magnitude and frequency were integrated into a mathematical model accounting for whole-body Aβ peptide kinetics. Validation of the developed model was performed through comparison with acute monomer levels in the blood serum of 15 service members exposed to repeated low-level blast while undergoing three-day weapons training. Amyloid precursor protein (APP) synthesis was assumed to be proportional to blast magnitude and additive effects within a window of recovery were applied to account for cumulative exposure. Aβ42 concentrations in the blood serum were predicted within 6.5 ± 5.2% on average, demonstrating model feasibility and biomarker sensitivity to blast. Outcomes discuss how modulation of patient-specific factors (age, weight, genetic factors, years of exposure, sleep) and pathophysiological factors (BBB permeability, amyloidogenic pathology, neuroinflammation) can reveal potential sources of variability in experimental data and be incorporated into the blast-dose BxK platform in future iterations. Advancements in model complexity accounting for sex-specific factors, weapon system, stress levels, risk of symptom onset, and pharmacological treatment strategies are anticipated to improve model calibration. Utilization of this blast-dose BxK model to identify drivers of pathophysiological mechanisms and predict chronic outcomes has the potential to transform bTBI diagnostic, prognostic, and therapeutic strategies.

Keywords: Aβ42; biomarker; blast; brain; diagnostics; modeling; serum.

Figure 1

Schematic of model development approach: (1) model, (2) integration, (3) validation, and (4) calibration.

Figure 2

Modeling mechanical damage ( $R_M$ ) and multi-phase biological responses ( $R_{B1}$ and $R_{B2}$ ) to blast exposure. (A) Biological responses to a blast magnitude below the injury threshold normalize to zero, representing full recovery. Time to peak $R_M$ was ~500 ms. (B) Biological responses to a blast magnitude above the injury threshold no longer normalize to zero, representing residual injury proportional to BOP magnitude. Figure annotations adapted from Blennow et al. (11). (C) If a repeated blast occurs within the recovery phase, the response was assumed to be additive.

Figure 3

Integration of blast-dose response model with Aβ biomarker kinetics. (A) The blast-dose biological responses $\left(R_B\right)$ drive APP synthesis in the amyloidogenic pathway, affecting Aβ monomer concentrations in the interstitial fluid (ISF). (B) Aβ peptide flux between tissue and blood was defined to include receptor-mediator endocytosis and bi-directional transport. (C) Biomarker concentrations in the serum were predicted using a PBPK model adapted from Bloomingdale et al. (44), incorporating exchange between brain, whole body tissue, lymph, and blood compartments.

Figure 4

(A) Weapons training timeline based on assumptions from Thangavelu et al. (24). (B) Blast-dose + biological response curves for four subject-specific cases. (C) Blast-Dose BxK model serum Aβ42 concentrations compared to experimental serum data for four subjects.

Figure 5

Absolute percent error of average pre-training and post-training model predictions for each person. Pre-training model predictions had significantly greater percent error compared to experimental data than post-training model predictions (**p < 0.01). Statistics were calculated using a two-tailed Wilcoxon paired t-test. Results are displayed as Mean ± SD.