Deep learning reveals how cells pull, buckle, and navigate fibrous environments

Cells in tissues navigate fibrous environments fundamentally differently than they do on flat substrates, but the establishment of cell forces in physiological fibrous settings remains poorly understood. Although factors such as the stiffness of the extracellular matrix (ECM) are known to drive behaviors, including cell motility on flat nonfibrous substrates, the interplay between fiber architecture and stiffness in fibrous ECM is not known. Here, we find that in fibrous environments, the directionality of mechanical forces overrides ECM stiffness as the primary regulator of contractility in migrating cells. Using an approach combining phase microscopy with deep learning to map forces in real time, termed deep learning-enabled live-cell fiber-force microscopy (DLFM), we reveal that when cells transition between anisotropic and isotropic stress fields, their contractility significantly drops despite encountering stiffer ECM, contrary to the behavior of cells on flat nonfibrous substrates. Unlike the peripheral adhesions observed on flat nonfibrous substrates, cells in fibrous matrices form force-generating adhesions throughout their body, stabilized by out-of-plane mechanical components unique to fiber geometry. Cells exhibit distinct force signatures during migration, division, and differentiation, with temporal signatures that predict stem cell fate. These findings, enabled by combining deep learning and the mechanics of cells and fibers, explain long-standing paradoxical behavior of cells navigating deformable fibrous environments, how they can pull and tug at them, and identify tension anisotropy as a master regulator of cell behavior, with implications for cancer invasion, tissue engineering, and regenerative medicine.