Quantifying stroma-tumor cell interactions in three-dimensional cell culture systems.

Background: In cancer research, studying cell motility is fundamental to investigate cancer invasion and drug resistance. In solid tumors constituted by a huge stroma component, such as pancreatic ductal adenocarcinoma (PDAC), the ability to quantify cell movement and interactions is mandatory to better understand the complex crosstalk between cancer and stroma cells. In PDAC, the pancreatic stellate cells (PSCs) are the principal source of dense fibrotic stroma: these closely interact with the PDAC cells to create a facilitating tumor microenvironment that supports local and distant tumor progression through the secretion and/or the response to a number of cytokines that globally increase cancer invasiveness; moreover, the dense tumor microenvironment contributes to resistance to chemotherapy and radiation therapy. Recent studies have established that targeting the stromal compartment in PDAC may lead to promising outcomes. Unfortunately, significant improvements in the overall survival of patients have not been realized in more than four decades, in part because of the lack of relevant preclinical models. Methods: To infer interactions existing between stromal and cancer cells in complex 3D environments, we developed a novel platform that combines time-lapse fluorescence microscopy, automated image detection, and extensive statistical analysis. To better mimic in situ PDAC, we created a hydrogel-based (Matrigel, collagen) model that allows cell tracking in 4 dimensions (x, y, z, t), not possible in 2D, using a mixed population of L3.6PL pancreatic cancer cells and PSCs at a ratio of 25% and 75%, respectively. To precisely identify the two populations and facilitate automatic cell detection, cells were transfected with GFP and mCherry vectors. Once in the 3D conformation, the cocultures were monitored through time-lapse confocal microscopy (CLSM) in controlled conditions, and the dataset containing the temporal evolution of the cells was processed by statistical tools. Results: Key to our approach was the generation of new machine learning-driven automated inferential protocols that resulted in the high-resolution imaging of the strength of the interactions between cells, as well as the potential presence of local chemokine gradients. In our ongoing studies, both commercial PDAC cell lines and pancreatic stellate cells were used to establish the model system. We are currently adopting the model to take full advantage of patient-derived PDAC cells established using the conditionally reprogrammed cells technique, mixed with patient-matched stellate cells. Since cell migration is a hallmark of cancer, we believe that this platform could be used as a reliable and reproducible approach for studying single-cell migration and invasion also in patient-derived models.