Sampling from single-cell observations to predict tumor cell growth in-vitro and in-vivo

// Alexander T. Pearson 1 , Patrick Ingram 2 , Shoumei Bai 3 , Patrick O'Hayer 4 , Jaehoon Chung 5 , Euisik Yoon 6 , Trachette Jackson 7, * and Ronald J. Buckanovich 3, * 1 Section of Hematology/Oncology, Department of Medicine, University of Chicago, Chicago, IL, USA 2 Department of Biomedical Engineering, University of Wisconsin, Madison, WI, USA 3 Magee-Womens Research Institute, University of Pittsburgh, Pittsburgh, PA, USA 4 University of Michigan School of Medicine, Ann Arbor, Michigan, MI, USA 5 Institute of Microelectronics, Science and Engineering Research Council of the Agency for Science, Technology and Research, Singapore 6 Department of Electrical Engineering and Computer Science, University of Michigan College of Engineering, Ann Arbor, MI, USA 7 Department of Applied Mathematics, University of Michigan, Michigan, Ann Arbor, MI, USA * Co-senior authorship Correspondence to: Ronald J. Buckanovich, email: buckanovichrj@mwri.magee.edu Keywords: cancer modeling; ovarian cancer; cancer stem cell; EGFL6; microfluidics Received: June 22, 2017 Accepted: October 16, 2017 Published: November 25, 2017 ABSTRACT Cancer stem-like cells (CSCs) are a topic of increasing importance in cancer research, but are difficult to study due to their rarity and ability to rapidly divide to produce non-self-cells. We developed a simple model to describe transitions between aldehyde dehydrogenase (ALDH) positive CSCs and ALDH(-) bulk ovarian cancer cells. Microfluidics device-isolated single cell experiments demonstrated that ALDH+ cells were more proliferative than ALDH(-) cells. Based on our model we used ALDH+ and ALDH(-) cell division and proliferation properties to develop an empiric sampling algorithm and predict growth rate and CSC proportion for both ovarian cancer cell line and primary ovarian cancer cells, in-vitro and in-vivo . In both cell line and primary ovarian cancer cells, the algorithm predictions demonstrated a high correlation with observed ovarian cancer cell proliferation and CSC proportion. High correlation was maintained even in the presence of the EGF-like domain multiple 6 (EGFL6), a growth factor which changes ALDH+ cell asymmetric division rates and thereby tumor growth rates. Thus, based on sampling from the heterogeneity of in-vitro cell growth and division characteristics of a few hundred single cells, the simple algorithm described here provides rapid and inexpensive means to generate predictions that correlate with in-vivo tumor growth.