AMIT: a novel algorithm for migration and interaction tracking for high-throughput analysis of phagocytosis assays

AMIT: a novel algorithm for migration and interaction tracking for high-throughput analysis of phagocytosis assays Susanne Brandes, Stefanie Dietrich, Kerstin Hünniger, Oliver Kurzai, Marc Thilo Figge Leibniz Institute for Natural Product Research and Infection Biology - Hans-Knöll Institute, Jena, Germany Research Group: Applied Systems Biology, E-Mail: susanne.br[email protected] Time-lapse microscopy is an important technique to study the dynamics of various biological processes. Imaging of biological systems combined with the analysis for functional, dynamical, and morphological aspects is required to increase our understanding of complex processes. In the context of interactions between human innate immune cells and pathogens, algorithms are required for automated high-throughput analysis of time-lapse microscopy videos of confrontation assays. We present an automated segmentation and tracking framework for analyzing phagocytosis assays of polymorphonuclear neutrophils (PMNs) confronted with C. glabrata. The algorithm is based on our previously developed framework for tracking of non-rigid cells in brightfield microscopy [1] and was extended by (i) a segmentation approach for fluorescently-labeled pathogens and (ii) a state-transition-model and a cross-linking procedure for track interactions between PMNs and fungal cells. The PMN segmentation approach yields a sensitivity of 99% and a precision of 95% in object detection and the segmentation approach for fluorescently-labeled fungal cells yields a sensitivity of 84% and a precision of 93%. The phagocytic activity of PMNs is quantified in terms of percentage of phagocytically active PMNs, number of phagocytosis events per PMN, frequency of phagocytosis events, and killing rate of phagocytosed fungal cells. The findings of our automated analysis approach are directly compared with manual and experimental analyses [2, 3] and yield high consistency. AMIT comprises an automated segmentation and tracking framework and paves the way towards quantitative high-throughput analysis of time-lapse microscopy videos of confrontation assays. Live-cell Imaging Data Multi-channel time-lapse microscopy videos of confrontation assays of PMNs and Candida glabrata. brightfield channel green fluorescence channel red fluorescence channel Segmentation Individual segmentation of different channel images. brightfield channel green fluorescence channel red fluorescence channel Brightfield images are segmented using a Gaussian-Mixture-Model (GMM) that is based on spatio-temporal image variances [1]. Fluorescence images are segmented using morphological operations and Canny edge detection (green fluorescence channel) or Ostu’s thresholding (red fluorescence channel). Additionally, fungal cell clusters are separated into their constituent cells using a method developed by Farhan et al., which detects concavity points in object contours [4]. fungal cell cluster split line estimation splitted fungal cell cluster Classification Identification of different cell types and PMN states. non-touching PMN (state 0) fungal cell cluster PMN touching a fungus (state 2) PMN with phagocytosed fungus (state 3) State 4: Interaction between PMN in state 0 and 3 State 5: Interaction between PMN in state 3 and a fungal cell cluster dead cell Interaction Tracking State-transition model. 0 start 2 34 5 fungi fungi PMN PMN fungi fungi The tracking of PMNs is based on the tracking algorithm developed in [1] and extended for the detection of interactions with fungal cells. The state-transition model for PMN states represents the basic principles of pathogen–phagocyte interactions. Results Quantitative analysis of cell track data reveal information about the growth of the fungal cell population as well as the phagocytic behavior of PMNs. ● ●●●●● ● ●●● ●●●●●●●●●●●●●●●●● ●● ●●●●●●●●●●●● ● ●● ●●●●●●●●●●●●●●●● 0 10 20 30 40 50 60 0.0 0.2 0.4 0.6 0.8 Time (min) Fungal growth (%) 1 2 Number of phagocytoses per PMN Proportion 0.0 0.2 0.4 0.6 0.8 1.0 ● ● 0−10 10−20 20−30 30−40 40−50 50−60 Time (min) Frequency of phagocytosis events 0.0 0.1 0.2 0.3 0.4 ● ● ● ● ● ● 1 2 3 4 5 6 7 8 Cluster size Frequency 0.0 0.1 0.2 0.3 0.4 0.5 0.6 ● ● ● ● ●●●● ● ● ● ● ●●●● Occurrence Phagocytosis References [1] S. Brandes, Z. Mokhtari, F. Essig, K. Hünniger, O. Kurzai, and M. T. Figge. Automated segmentation and tracking of nonrigid objects in time-lapse microscopy videos of polymorphonuclear neutrophils. Medical Image Analysis, 20(1):34–51, 2015. [2] S. Duggan, F. Essig, K. Hünniger, Z. Mokhtari, L. Bauer, T. Lehnert, S. Brandes, A. Häder, I. D. Jacobsen, R. Martin, M. T. Figge, and O. Kurzai. Neutrophil activation by Candida glabrata but not Candida albicans promotes fungal uptake by monocytes. Cellular Microbiology, 17(May):1259–1276, 2015. [3] F. Essig, K. Hünniger, S. Dietrich, M. T. Figge, and O. Kurzai. Human Neutrophils dump Candida glabrata after intracellular killing. Fungal Genetics and Biology, 2015. [4] M. Farhan, O. Yli-Harja, and A. Niemistö. A novel method for splitting clumps of convex objects incorporating image intensity and using rectangular window-based concavity point-pair search. Pattern Recognition, 46(3):741–751, 2013. Acknowledgement This work was supported by the Leibniz Association.