SP05. A Novel, Patient-specific, Organotypic 3D Platform for Modeling Cell Behavior in the Setting of Malignancy

Abstract

PURPOSE: With the advent of personalized cancer treatments, patient-specific in vitro models of the breast malignancies are critical for modeling the complex in vivo milieu and to better assess therapeutic efficacy. A primary challenge to developing such platforms is isolating and successfully co-culturing the many primary cell types that constitute the tumor microenvironment. In addition, optimizing for cost-effectiveness of such platforms is critical to making the technology translational. Using patient-derived cells and machine learning-based image analysis, we have developed a low-cost 3D biomimetic platform that allows for the study of cell behavior in a highly organotypic model of breast malignancy. METHODS: Stromal vascular fraction, organoids and mature adipocytes were isolated from breast tissue obtained from healthy female patients undergoing reduction mammoplasty. Isolated cells were embedded in non-ribosylated collagen, creating a biomimetic milieu that approximates the patient-specific in vivo cellular environment (“biomimetic collagen”). 3D collagen constructs consisting of a bottom layer of plain collagen or collagen embedded with RFP-tagged MDA-231 tumor cells followed by a layer of plain or biomimetic collagen were built in a 96-well plate. GFP-tagged human umbilical vascular endothelial (HUVEC) cells were plated in a monolayer layer on top of the collagen to mimic the endothelial barrier. Constructs were imaged with confocal microscopy on days 0 and 7 to assess for endothelial cell organization and migration. Image analysis was completed in Imaris. ANOVA tests for statistical significance were completed in R Studio. RESULTS: Confocal microscopy at day 0 revealed ~800μm thick constructs with distinct tumor and non-tumor collagen layers and a robust HUVEC top monolayer. Confocal images at day 7 were notable organization of the HUVEC layer into 2D tubular networks with some HUVEC below the top monolater indicating migration through the collagen matrix. Our machine-learning based approach was able to reliably and reproducibly parse HUVEC cells from the surrounding cells and collagen, allowing for quantification of cell migration in the Z-axis. Statistical analysis revealed a significant difference in migration across treatment groups (p < 0.01), with endothelial cells in the tumor and biomimetic collagen platform migrating 20-40 microns further than endothelial cells in the other construct designs, a finding retained on intergroup pairwise comparison (p < 0.01). CONCLUSION: This biomimetic platform design and image analysis allows for reliable, reproducible study of discrete endothelial cell interactions with the tumor microenvironment. The finding of enhanced HUVEC migration in the platform with both tumor and biomimetic cells is consistent with the hypothesis that stromal cells, adipocytes and malignant cells enhance the angiogenic capacity of endothelial cells. Importantly, this utilization of patient-derived stromal vascular fraction and mature adipocytes in this platform makes it inherently personalized, a vital component of any future clinically-relevant in vitro cancer platform.