MULTISCALE ANALYSIS OF COLLECTIVE CELL MOTION

Abstract

Collective cell migration plays an important role in a variety of biological processes such as tissue formation, wound healing, cancer progression, and immune response. However, it is still a challenge to quantitatively correlate emergent behavior to individual cell identity, genetic background and environment surrounding the cells. In this thesis, I present a multiscale analysis framework that extracts and interprets motion phenotypes from both in vitro and in vivo systems, integrating computational approaches with biological insight.I first developed an AI-based label-free tracking method to analyze cell motion in epithelial monolayers, using phase contrast imaging. From these tracks, we calculated motion metrics such as D2min, which captures local non-affine rearrangements, and affine deformation, which reflects coordinated motion. Using these metrics to examine a panel of breast epithelial and cancer cell lines, we observed that more invasive cells exhibited greater D2min and less directional affine motion. These metrics can be used to accurately separate the nontumorigenic, tumorigenic, and metastatic behaviors. We then explored how physical cues like nano-ridged substrates affect migration. While topography enhanced speed and directionality, it had a limited effect on mixing. Rather, we found that cell identity, in our case, KRas and PTEN mutations, dominates the response, overriding environmental influence and maintaining segregation despite changes in substrate. To apply this analysis to live tissues, we performed 3D optical flow on intravital imaging data of immune cells in mouse tumors. This method captured collective myeloid motion near tumor lesions and revealed dynamic motion hotspots. We tracked these regions over time and observed that early-stage motion patterns may relate to whether lesions progress or regress, hinting at a predictive role for immune behavior. Overall, these studies show that collective cell motion is influenced by genetic programs inside the cells and also by signals from the environment. This thesis provides a scalable and label-free method to quantifying motion behavior in systems across scales, offering tools for both basic biology and translational research.