Parallelized lensfree cell migration screening

PARALLELIZED LENSFREE CELL MIGRATION SCREENING I. Ghorbel1,2,3, F. Kermarrec1,2,3, B. Sartor1, X. Gidrol1,2,3 and V. Haguet1,2,3* 1 CEA iRTSV, FRANCE, INSERM U1038, FRANCE and 3 University Grenoble-Alpes, FRANCE 2 ABSTRACT An array of 12×8 CMOS image sensors was produced for high-throughput holographic video microscopy of cell populations cultured in a 96-well microtiter plate. Wound healing assays were performed in microtiter plates to screen for 2D migration of cell lines. Image processing based on global segmentation algorithms was programmed to uncover dynamics of wound closures. Motilities of prostate cell lines having the same origin were investigated using parallelized lensfree video microscopy and image analysis. N-methyl-N-nitrosurea-transformed tumor cell lines derived from the healthy RWPE-1 prostate cell line demonstrate higher migration capabilities, independent from the degree of invasiveness, than the original healthy cells. KEYWORDS: Parallelized lensfree imaging, Wound Healing Assay, Microtiter plate, Prostate cells INTRODUCTION In previous work, we have developed lensfree imaging of cell cultures to achieve video microscopy of cell growth and motility directly in a CO2 incubator [1,2]. Images of cell populations cultured in 35mm Petri dishes on a CCD image sensor were typically recorded for 3 days every 4 to 10 minutes to build movies of cell proliferation and measure confluence over time. Recently, we expanded this approach to phenotypic screening and achieved high-throughput lensfree imaging in transparent-bottomed 96-well microtiter plates using an array of 96 CMOS image sensors (Figure 1) [3,4]. Home-made housing was designed and fabricated to fix the printed circuit board holding the 96 image sensors and perform in situ microscopy of every cell culture in the multiwell plate. Parallelized quantitative image acquisitions of cell dynamics were exploited to investigate cell adhesion and proliferation, optimize cell culture conditions and detect previously unknown spontaneous 3D cell aggregation [3]. Figure 1: Array of 12×8 CMOS image sensors (left side of the printed circuit board) used to capture images of cell cultures in a 96-well microtiter plate. Here, we present high-throughput visualization and quantification of cell motility based on 96 wound healing assays performed simultaneously in the microtiter plate. In addition, we developed automated image processing of the 96 image series supplied by the video microscope. Parallelized time-lapse image acquisition and dedicated image analysis were applied to determine and compare cell migration capabilities of the healthy RWPE-1 prostate cells and 3 derived prostate cancer cell lines. EXPERIMENTAL AND RESULTS The wounds were performed in a single step using a wounding replicator with 96 pins. The resulting wounds are approximately 500 µm large and can be easily visualized using the 2.8×2.1 mm2 imaging area and 1.75 µm pixel size of the CMOS image sensors. Figure 2 shows a global view of the 96 wounds made in confluent prostate cell cultures and imaged by parallelized lensfree microscopy. a Figure 2: Global view at t0+30mn of the 96 “verticallyoriented” wounds made in 4 prostate cell cultures. b Figure 3: Two steps of the wound segmentation algorithm: (a) k-means/Markov random field process; (b) parallel double snakes algorithm [5]. Every culture condition for a given cell line was reproduced in 2 microtiter plates and in 8 wells in every microtiter plate. Dynamics of the wound closures were typically recorded for 24 hours. The two edges of the wounds were detected on the images by global segmentation algorithms based on Markov random fields and active contours, followed by a parallel double snake (Figure 3) [5]. Accuracy and robustness of our image processing technique was successfully compared with manual segmentations by seven cell biology experts. We verified that the index of similarity (S), which is sensitive to variations in shape, size and position, is in all cases greater than 0.98, which denotes a strong agreement between automatic and manual segmentations [6]. The edge-to-edge distance and the area uncovered by the cells were measured at every time point for the entire multi-well array (Figure 4). The decreases of wound width and uncovered area were compared relatively to the initial values determined for every well few minutes after the wounds were made, i.e., when the cells are still in latent phase resulting from stress made by mechanical detachment. (a) (b) t0+30mn t0+3h (c) (d) t0+7h10 t0+10h Figure 4: Evolution of a wound made in confluent WPE1-NA22 cells, normalized relatively to the initial width (also called thickness) of the wound. Wound closure kinetics of the RWPE-1 prostate cell line from a healthy donor was compared with the ones by three N-methyl-N-nitrosurea (MNU)-induced cancer cell lines derived from RWPE-1 and representing different grades of tumor progression. For the 3 cancer cell lines, wound closure dynamics shows a sharp decrease of the wound width until the two edges of the wound meet (Figure 4). Most rapid cells fully invaded the initial wound areas in 10 hours, so this time point was selected to be the end point for the wound healing assay experiment. Migration of the WPE1-NA22, WPE1-NB11 and WPE1-NB26 tumor cells, of increasing invasiveness, appears to be comparable and far superior to the one of healthy RWPE-1 cells (Table 1). Unexpectedly, the less transformed cells, WPE1-NA22, already have a strongly modified motility pattern compared to the original RWPE-1 cells. Table 1: Wound closure kinetics of the RWPE-1 prostate cell line and 3 MNU-derived cancer cell lines. Cell lines Ratio of wound closure after 10 h RWPE-1 WPE1-NA22 WPE1-NB11 WPE1-NB26 14% 100% 78% 98% CONCLUSION Parallelized holographic microscopy and automated image quantification were developed to support phenotypic screening in 96-well microtiter plates. Wound healing assays were carried out to quantify migration of prostate cell lines. MNU transformation of RWPE-1 prostate cells is shown to strongly increase cellular motility, even for the less transformed cells WPE1-NA22. ACKNOWLEDGEMENTS Financial support was provided by the National Research Agency (grant ANR-12-EMMA-0051-01). REFERENCES [1] M. Gabriel, V. Haguet, N. Picollet-D’hahan, M. Block, B. Fouqué and F. Chatelain, “Inexpensive integrated cell imaging device”, Proc. MicroTAS, Oct. 12-16, 2008, San Diego, USA, vol. 2, pp. 1600-1602. [2] M. Gabriel, N. Picollet-D’hahan, M. Block and V. Haguet, “Monitoring adherent cells by contact imaging”, Proc. MicroTAS, Nov. 1-5, 2009, Jeju, South Korea, vol. 1, pp. 278-280. [3] V. Haguet, P. Obeïd, R. Griffin, D. Freida, L. Guyon and X. Gidrol, “Time-lapse screening by parallelized lensfree imaging”, Proc. MicroTAS, pp. 1740-1742, Freiburg, Germany, Oct. 27-31, 2013. [4] I. Ghorbel, N. Bertacchi, X. Gidrol and V. Haguet, “Parallelized contact imaging and automated analysis of cell migration dynamics”, Proc. 37th International Meeting of the German Society for Cell Biology, Regensburg, Germany, March 18-21, 2014. [5] I. Ghorbel, F. Rossant, I. Bloch and M. Paques, “Modeling a parallelism constraint in active contours. Application to the segmentation of eye vessels and retinal layers”, Proc. 18th IEEE International Conference on Image Processing (ICIP), pp. 445-448, Brussels, Belgium, Sept. 11-14, 2011. [6] A. Zijdenbos, B. Dawant, R. Margolin and A. Palmer, “Morphometric analysis of white matter lesions in MR images: method and validation”, IEEE Trans. Med. Imag., 13(4), 716–724, 1994. CONTACT * V. Haguet; phone: +33-4-38-78-23-86; [email protected] View publication stats