| Summary: | This thesis presents the design and validation of advanced image analysis and machine learning tools for the quantitative analysis of three-dimensional (3D) lung cancer cell migration. These tools were used to study the role that the composition and the morphological and mechanical properties of the microenvironment have in cancer cell migration.
Cell migration experiments were performed using in vitro cellular models consisting of custom-made microfluidic devices, which provide a high level of control of the properties of the microenvironment, offer optimal optical properties for microscopic observation and ensure appropriate use of reagents due to their microscopic dimensions.
Varying migration environmental conditions were simulated by embedding the cells in biomimetic hydrogels of different composition and, thus, with different morphological and mechanical properties. These hydrogels act as the 3D extracellular matrix (ECM) of the cells. Namely, we used hydrogels made of pure collagen type I to mimic normal connective tissue and hydrogels made of a mixed composition of collagen and Matrigel to simulate the disorganized basement membrane at the leading edge of cancer invasion.
The migration experiments performed within the hydrogels were done with and without serum stimulation to determine the role of hydrogel extrinsic and intrinsic growth factors in cell migration. Furthermore, migration experiments were performed with and without blocking cellular integrins to determine the role of the cell-to-ECM interactions in cell migration, and with and without metalloproteinase (MMP) inhibitors to understand the role of ECM in cell migration.
Migration experiments were performed at two resolution levels. First, low-resolution experiments were performed to determine globally the effect of the parameters analyzed in the dynamics of cancer cell migration. This required detecting and tracking a large number of cancer cells, imaged at low magnification (10×) using widefield phase-contrast and fluorescence microscopy. Second, high magnification (63×) migration experiments were used to study the morphology of migrating cancer cells, by measuring the number, length and life time of cell protrusions produced during the migration. This required the use of laser scanning confocal microscopy.
The microscopy time-lapse sequences generated in both types of experiments were analyzed using novel image processing and machine learning techniques in order to obtain robust quantifiable metrics of cell migration dynamics and migrating phenotype in 3D environments.
This thesis describes the experiments performed along with the image processing and computational algorithms used to analyze cell migration both in low- and high-magnification image data. We provide the source code of the developed analysis software, for its use and future algorithmic improvement by the research community.
|
|---|