Abstract:Mechanical signals drive neural cell growth and migration during development and regeneration. Although most cells migrate towards stiffer substrates, neurons can migrate in the opposite direction. We hypothesized that neural growth cones, known to affect stiffness sensing, are stiffness sensors as well, and tested this hypothesis via a mathematical model. The model integrates interactions between axons, growth cones and substrate through a two-dimensional molecular clutch model of filopodial dynamics and a viscoelastic model of active axon contraction. The modeling results predict that different types of neurons should have distinct mechanosensing behaviors, with dorsal root ganglion cells behaving differently than hippocampal neurons. Also, substrate stiffness can guide growth cone motion through altering the traction force generated in the growth cone. The model provides the first explanation of how stiffness can be sensed and provides mechanistic insight into how axon-growth cone-substrate interactions drive neural regeneration and repair. Since existing models failed to address the interactions between axons, growth cones and ECM stiffness in neural mechanosensing, we constructed a mathematical model of growth cone based on the molecular clutch framework of Odde et al., which treats focal adhesion proteins and the substratum as linear springs loaded by a cytoskeletal tension, to investigate how substrate stiffness affects growth cone dynamics and axonal retraction. Briefly, a group of i filopodia, each modeled as a single molecular clutch module, are attached to an axon that is modeled as a contractile element in parallel with a standard linear viscoelastic solid. Each molecular clutch module contains n(i) adhesion clutches (linear springs of stiffness k(i)) and n(i) corresponding substrate clutches (linear springs with stiffness E-i). Substrate clutches are distributed randomly, and have values E-1 <= E-i <= E-2 that vary linearly with position as a first order approximation to the spatially graded, nonlinear mechanical properties of ECM. Adhesion clutches bind to substrate clutches with a rate of k(on), forming a force transmission pathway from intracellular cytoskeleton, where contraction arises from inward actin flow, to the extracellular substrate. The unbinding rate of k(off,i) varies with the force F-i in the connected clutch according to the Bell model. We first studied the different behaviors between neurons from dorsal root ganglion and hippocampus respectively and found that the expression of adhesion proteins on growth cone is vital for the generation of traction force. Since traction force is determinant to the axon dynamic processes including elongation and retraction, the expression of adhesion proteins thus plays a vital role in neuron mechanical behaviors. Then we tried to predict how the axon responds to substrate with different stiffness by combining the growth cone model with a viscoelastic contractile axon body model. The combined model illustrated how the contraction of myosin molecular motors decreases as the axon deformation rate goes down. The results further showed that low motor contraction can cause the axon retraction on high-stiffness substrate while much higher contraction force is needed for the axon retraction on low-stiffness substrate. The predictions conform to previous experimental observations in the literature and indicate that the increase of motor contraction leads to the axon retraction on the substrate with the same stiffness, providing the possible reasons that neurons usually migrate towards the low-stiffness area instead of durotaxis seen in other cell types. Altogether, our model results suggest that growth cone filopodial dynamics, including dynamics of turning, enlargement and contraction, can explain how substrate stiffness affects neuronal mechanosensing, and shed light on how the nervous system might develop to distinguish and reach target tissues and cells by mechanosensing.

Neural networks for biomechanics

Zyxin is all you need: machine learning adherent cell mechanics

Machine learning interpretable models of cell mechanics from protein images

Artificial Neural Network-Aided Computational Approach for Mechanophenotyping of Biological Cells Using Atomic Force Microscopy

Machine learning traction force maps of cell monolayers

Real-time calibration-free musculotendon kinematics for neuromusculoskeletal models

G2Φnet: Relating genotype and biomechanical phenotype of tissues with deep learning

Predicting the Mechanical Properties of Biopolymer Gels Using Neural Networks Trained on Discrete Fiber Network Data

Feed forward artificial neural network to predict contact force at medial knee joint: Application to gait modification

Predicting Knee Joint Contact Forces During Normal Walking Using Kinematic Inputs with A Long-Short Term Neural Network

A Neural Network Model for Efficient Musculoskeletal-Driven Skin Deformation

Knee-Loading Predictions with Neural Networks Improve Finite Element Modeling Classifications of Knee Osteoarthritis: Data from the Osteoarthritis Initiative

Automated personalization of biomechanical knee model

Deep learning strategy for small dataset from atomic force microscopy mechano-imaging on macrophages phenotypes

A trained neural network model accurately predicts Achilles tendon stress during walking and running based on shear wave propagation

Neural Computing in Mechanics

A generic physics-informed neural network-based constitutive model for soft biological tissues

Prediction of lower limb joint angles and moments during gait using artificial neural networks

A Mechanical Model of Neuron Axons Considering Extracellular Matrix Stiffness

Forced peeling and relaxation of neurite governed by rate-dependent adhesion and cellular viscoelasticity

A finite element model of an osteoblast to quantify the transduction of exogenous forces to cellular components