S. Weber(1,2), M.L. Fernandez-Cachon(1,2), R.F. Murphy(1,3), M. Boerries(1,2) and H. Busch(1,2)
(1) School of Life Sciences, Freiburg Institute for Advanced Studies, Albert-Ludwigs-University of Freiburg, Germany; (2) Center for Biological Systems Analysis, Albert-Ludwigs-University of Freiburg, Germany; (3) Computational Biology, Carnegie Mellon University, Pittsburgh, USA
Objectives: Neuronal cell fate decisions are an active field of research and the detection of the cell phenotype differentiation, i.e. the outgrowth of neurites, is essential. We have developed the first classification approach capable of detecting the differentiated phenotype by using phase contrast imaging of live cell populations only [1]. In contrast to standard techniques, this represents a non-destructive and interference free sampling, which enables longitudinal studies of a cell population. Hence, this allows for the first time a mixed effect analysis in this context. The main motivation is to gain a more detailed understanding of cell differentiation kinetics.
Methods: As a model system we use PC12 cells [2], which undergo cell differentiation under treatment with nerve growth factor (NGF). We monitored control (CTL) cells without treatment and cells treated with NGF on dedicated wells. We recorded 300 images each day per well in a line-wise spatial recording pattern. As this pattern remains constant over the observation time, we approximately monitor the same cell population each day in each image. Image feature extraction is performed in MATLAB and further statistical analysis is done with Monolix 4.1.2 [3].
Results: The key step was to identify the image features, which detect the differentiated phenotype and are suitable for further modeling. First, cells or cell clumps are detected in each image in a segmentation step. Then for each cell segment the convex hull covering area and the cell area itself is determined. The convex hull covering grows faster for cells which outgrow neurites than for cells which only proliferate. Hence, we use the convex hull area as the readout parameter for the differentiation and the cell area as the readout parameter for the cell proliferation. Both processes are subjected to limited resources and we hence model these by the Verhulst equation. Main results are that (i) the treatment is statistically highly significant when comparing the reduced to the full model in both cases of differentiation and proliferation, (ii) extracted growth rate constants stay in physiologically reasonable ranges and (iii) NGF treatment leads to a doubling of the growth rate of the convex hull area.
Conclusions: We have modeled neuronal cell proliferation and differentiation kinetics by means of a limited growth mixed effect model. We were able to extract growth rate constants for cell proliferation and characterize cell differentiation kinetics. We expect this approach to be of great value for the study of neuronal cell differentiation by means of further treatment combinations.
References:
[1] S. Weber, M.L. Fernandez-Cachon, B. Offermann, R.F. Murphy, M. Boerries, and H. Busch, (2012), Classification of neuronal cell differentiation from phase-contrast imaging, To be submitted
[2] L.A. Greene and A.S. Tischler (1976), Establishment of a noradrenergic clonal line of rat adrenal pheochromocytoma cells which respond to nerve growth factor. PNAS, 73(7), 2424 –2428.
[3] Kuhn, E., and Lavielle, M. Maximum likelihood estimation in nonlinear mixed effects models. Computational Statistics and Data Analysis 49, 4 (2005), 1020-1038.
Reference: PAGE 21 (2012) Abstr 2537 [www.page-meeting.org/?abstract=2537]
Poster: Other Modelling Applications