Learning Epithelial Elasticity via Local Tension Remodeling.

Biological materials, like epithelial tissues, exhibit remarkable adaptability to mechanical stresses, dynamically remodeling their structure in response to external and internal forces. A key challenge is understanding how these tissues store a memory of past mechanical stimuli. Here, we investigate this memory using an active Vertex Model of epithelial sheets incorporating a local, mechanosensitive tension-remodeling rule where junctional tension updates depend on strain, acting as a slow, history-dependent variable. We demonstrate three hallmark mechanical consequences of this memory mechanism. First, a localized, short contractile cue permanently reprograms the global shear modulus, with the direction of change (stiffening or softening) controlled by the tension remodeling rate. Second, the tissue stores a long-range mechanical memory: a prior stimulus at one site modulates the tissue's response to a subsequent, distant stimulus, mediated by coupling across the entire junctional network. Finally, we show that simple cyclic bulk deformation acts as a training protocol that autonomously tunes the tissue's constitutive properties, including programming the Poisson ratio to auxetic (negative) values. These findings position epithelial mechanics within the framework of unsupervised physical learning, identifying the mechanosensitive remodeling rates as powerful control parameters for designing programmable tissue-scale rheology.