Learning epithelial elasticity via local tension remodeling
- 1School of Physics, Georgia Institute of Technology , 837 State Street NW, 30332, GA ,
Abstract
Biological materials such as epithelial tissues continuously remodel their internal structure in response to mechanical forces, but whether this adaptation amounts to a genuine form of physical learning has remained unclear. Here we show that an active vertex model of epithelial sheets, equipped with a single local, mechanosensitive tensionremodeling rule, exhibits the three canonical hallmarks of learning matter. First, at the scale of a single junction, repeated contractile pulses produce habituation: a diminishing, history-dependent response that encodes a memory of prior stimulation. Second, at the tissue scale, a localized stimulus writes a long-range mechanical memory that biases the response to a later, distant stimulus. We derive a continuum theory showing that this memory is transmitted as a system-wide elastic bias and predict its dependence on tissue size and remodeling rate. Third, under cyclic bulk driving, the tissue can be trained to a target mechanical state, autonomously tuning its Poisson ratio to auxetic (negative) values through directed remodeling rather than designed geometry. We further derive an effective viscoelastic memory kernel that distinguishes this adaptive tissue from a passive vertex model. Together, these results identify a candidate mechanism for physical learning in epithelia, in which the ratio of contraction to extension remodeling rates acts as the central control parameter for programmable, history-dependent tissue-scale rheology.
Related
- Rotational elastography: six degree-of-freedom seismology in soft tissues measures local elasticity with diffuse fields
- Amino Acid Remodeling Precedes Malignant Transformation in Oral Epithelial Cells
- Spatial transcriptomics reveals epithelial-immune remodeling preceding malignant transformation of oral premalignant lesions
- Campylobacter jejuni Infection Is Associated With Cell-Cycle, Redox, and Metabolic Remodeling of Human Intestinal Epithelial Cells at Single-Cell Resolution
- SIM2s coordinates programmed mitophagy with respiratory chain supercomplex remodeling during mammary epithelial differentiation
BibTeX
@article{arzash2026learning,
title = {{Learning epithelial elasticity via local tension remodeling}},
author = {Sadjad Arzash and Shiladitya Banerjee},
journal = {PNAS Nexus},
year = {2026},
month = oct,
doi = {10.1093/pnasnexus/pgag367},
url = {https://academic.oup.com/pnasnexus/advance-article/doi/10.1093/pnasnexus/pgag367/8997936},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Learning epithelial elasticity via local tension remodeling
AU - Sadjad Arzash
AU - Shiladitya Banerjee
PY - 2026
DA - 2026/10/25
JO - PNAS Nexus
DO - 10.1093/pnasnexus/pgag367
UR - https://academic.oup.com/pnasnexus/advance-article/doi/10.1093/pnasnexus/pgag367/8997936
AB - Biological materials such as epithelial tissues continuously remodel their internal structure in response to mechanical forces, but whether this adaptation amounts to a genuine form of physical learning has remained unclear. Here we show that an active vertex model of epithelial sheets, equipped with a single local, mechanosensitive tensionremodeling rule, exhibits the three canonical hallmarks of learning matter. First, at the scale of a single junction, repeated contractile pulses produce habituation: a diminishing, history-dependent response that encodes a memory of prior stimulation. Second, at the tissue scale, a localized stimulus writes a long-range mechanical memory that biases the response to a later, distant stimulus. We derive a continuum theory showing that this memory is transmitted as a system-wide elastic bias and predict its dependence on tissue size and remodeling rate. Third, under cyclic bulk driving, the tissue can be trained to a target mechanical state, autonomously tuning its Poisson ratio to auxetic (negative) values through directed remodeling rather than designed geometry. We further derive an effective viscoelastic memory kernel that distinguishes this adaptive tissue from a passive vertex model. Together, these results identify a candidate mechanism for physical learning in epithelia, in which the ratio of contraction to extension remodeling rates acts as the central control parameter for programmable, history-dependent tissue-scale rheology.
SN - 2752-6542
ER -