Deep Learning Enabled Label-free Cell Force Computation in Deformable Fibrous Environments
Cells move within tissues by pulling on and reshaping their fibrous surroundings. Measuring the associated forces has been a fundamental challenge in cell biology. Here, we develop deep-learning-enabled live-cell fiber-force microscopy (DLFM), which computes forces produced by living cells in real time as they interact with tissue-like fiber networks. DLFM combines basic phase microscopy with novel deep learning to simultaneously track cell movement and fiber deformation without disruptive fluorescent labels or chemical modifications. This allowed us to measure forces in real-time situations that were previously impossible to study, revealing an intricate mechanical landscape: cells generate ten-fold changes in force as they change shape during migration, create force-dipoles during cell-cell interactions, and dramatically alter their force patterns during stem cell differentiation. Through integrated experiments and mathematical modeling, we discovered that cells in fibrous environments form force-generating adhesions throughout their body, strikingly different from the edge-only adhesions seen in traditional petri dish experiments. Results clarify cytoskeletal pathways by which cells adapt force-generating machinery to navigate the fibrous architecture of tissues.