Solving multicomponent reaction-transport with coupled cellular trajectories and data-driven cellular activation models
Yichen Lu, Mei Yan Lee, Talid Sinno, Scott L. Diamond
Institute for Medicine and Engineering, Department of Chemical and Biomolecular Engineering, University of Pennsylvania, Philadelphia PA 19104
Introduction: The release of autocrinic factors by cells often triggers autocatalytic surface reactions, all of which in the presence of diffusive and transport may further strengthen the signal and showcase interesting phenomena of cellular growth. Such situations include: blood clotting/bleeding under flow, biofilm formation, and angiogenesis/growth factor/tumor interactions. The problem of blood coagulation involves a series of autocatalytic agonists. When blood flows over disrupted vessel surfaces, activating platelets release ADP and thromboxane characterized by mobilized calcium level, while their surface becomes catalytic for thrombin production via phosphatidylserine exposure. Thrombin further facilitates clot growth by triggering fibrin production which strengthens clot mass.
Materials and Methods: Platelet calcium mobilization was measured by pairwise agonist scanning (PAS) of responses to single and pairwise delivery of low, medium, and high doses of: ADP, thromboxane mimetic (U46619), collagen mimetic (convulxin) and thrombin. A neural network model was trained (Lee et al. PLoS Comput Biol, 2015) with the PAS data for use in multiscale simulations. A multiscale model of clotting under flow (Flamm et al. Blood, 2012) was expanded for enhanced speed and accuracy. An aggregate remodeling was applied to growing clots to achieve more realistic shapes. A particle event-driven triangular element remeshing module was used for efficient and adaptive finite element method (FEM) solution of soluble concentrations. Thrombin was generated at the injury site and its transport was calculated with FEM. An activation/aggregation model involving ADP, thromboxane and thrombin was developed utilizing the calcium mobilization data from an averaged 120 neural network response.
Results and Discussion: The remodeling scheme was implemented to allow for instantaneous local translocation and rolling of single platelets and was shown to make more physiologically consistent prediction of the clot morphology. The adaptive meshing scheme was applied to provide unstructured triangular meshes that vary along with the modulating clot contour so that better resolution and efficiency were achieved for calculation of ADP/TXA2 release and thrombin transport. The activation/aggregation model characterized by the experimentally trained neural network response was able to predict platelet deposition dynamics for whole blood flowing over collagen at 200 s-1 wall shear rate consistent with in-silico measurements. This approach allows efficient multiscale calculations, from the single cell to the tissue level, when the cellular ensemble has evolving structure driven by coupled species transport and prevailing hemodynamics.
Conclusions: The computer-aided multiscale approach has proved useful for mimicking the growth of thrombus under flow condition and making reasonable prediction on platelet morphology and activation. More involved biological complexity can be added to the current model to study various aspects of hemostasis under different conditions.
References:
- Lee MY, Diamond SL (2015) A Human Platelet Calcium Calculator Trained by Pairwise Agonist Scanning. PLoS Comput Biol 11(2): e1004118.
- Flamm M, et al. Multiscale prediction of patient-specific platelet function under flow. Blood. 2012;120:190–198