Abstract:

Tissues stiffen in aging, fibrosis, cancer, and injury, yet how these physical changes are recorded inside the nucleus has remained poorly understood. This dissertation establishes that matrix stiffness is not merely sensed at the cell surface and relayed biochemically inward. It is encoded directly in chromatin architecture, accumulated over time, and expressed as distinct cellular phenotypes depending on the duration and history of mechanical exposure.

The central argument is that chromatin behaves as an active mechanochemical material: its structure is maintained by a competition between physical coarsening and epigenetic reaction dynamics, and this competition makes it responsive, historical, and ultimately vulnerable to sustained mechanical perturbation. I show that this encoding unfolds across three timescales: minutes-to-hours, days-to-weeks, and months, each governed by a distinct mechanism and each building on the last.

At the shortest timescale, stiffness acts as an architectural perturbation. A sequencing-informed active polymer model reveals that nanoscale chromatin domains arise from competition between physical coarsening and acetylation- and methylation-associated reactions. This competition produces finite heterochromatin-rich domains whose boundaries are mechanically and biochemically labile, predicted by the model and confirmed by STORM imaging, Hi-C, RNA-seq, pharmacological perturbations, and substrate-stiffness experiments. Domain boundaries are the primary sites where mechanical inputs become architectural and transcriptional change.

This remodeling framework motivates an inverse problem: if spatially varying regulatory epigenetic fields shape chromatin architecture, can those fields be inferred from super-resolution images alone? I address this by developing mollifier-based physics-informed neural networks that recover hidden reaction fields from noisy super-resolution chromatin images, connecting biophysical modeling, microscopy, and machine learning into a unified inference framework.

At the days-to-weeks timescale, stiffness becomes memory. In human mesenchymal stem cells, prolonged stiff-matrix exposure produces asymmetric mechanical memory: stiff-primed cells retain stiff-like chromatin organization and transcriptional programs after return to compliant substrates, whereas soft-primed cells are=] readily overwritten by subsequent stiff culture. This asymmetry reflects impaired recovery of H3K9me3-marked heterochromatin rather than simple persistence of a stiffness-induced transcriptional state, a distinction with direct implications for how mechanical history is stored and potentially reversed.

At the months timescale, failed heterochromatin recovery becomes nuclear vulnerability. Chronic stiff culture accelerates DNA damage accumulation, nuclear envelope remodeling, loss of chromatin architecture and replication programs, reduced differentiation potential, and a senescence-like transcriptional trajectory. Chromatin-opening perturbations accelerate this progression; chromatin-protective interventions suppress it, supporting a model in which sustained mechanical load and heterochromatin erosion cooperate to deplete the nucleus of its buffering capacity.

Together, these results define a temporal hierarchy for nuclear mechanobiology. Stiffness is first encoded as nanoscale chromatin remodeling, then stabilized as asymmetric mechanical memory through impaired heterochromatin rebuilding and finally expressed as mechano-senescence as buffering capacity is exhausted. The nucleus is not only a regulator of cell identity. It is a recorder of mechanical experience, and the fidelity of that recording determines whether a cell adapts, remembers, or ages.