Date Published: June 3, 2019
Publisher: Public Library of Science
Author(s): Elaheh Alizadeh, Wenlong Xu, Jordan Castle, Jacqueline Foss, Ashok Prasad, Steven M. Abel.
A number of recent studies have shown that cell shape and cytoskeletal texture can be used as sensitive readouts of the physiological state of the cell. However, utilization of this information requires the development of quantitative measures that can describe relevant aspects of cell shape. In this paper we develop a toolbox, TISMorph, that calculates a set of quantitative measures to address this need. Some of the measures introduced here have been used previously, while others are new and have desirable properties for shape and texture quantification of cells. These measures, broadly classifiable into the categories of textural, irregularity and spreading measures, are tested by using them to discriminate between osteosarcoma cell lines treated with different cytoskeletal drugs. We find that even though specific classification tasks often rely on a few measures, these are not the same between all classification tasks, thus requiring the use of the entire suite of measures for classification and discrimination. We provide detailed descriptions of the measures, as well as the TISMorph package to implement them. Quantitative morphological measures that capture different aspects of cell morphology will help enhance large-scale image-based quantitative analysis, which is emerging as a new field of biological data.
The shape of a cell spread on a substrate is determined by the balance between the internal and external forces exerted on the cell boundary. The cell exerts forces and responds to external forces, from the extra-cellular matrix (ECM) or from neighboring cells, with the help of molecular motors and the cellular cytoskeleton, which is thus the ultimate determinant of cell shape [1, 2]. The cytoskeleton is a complex network, made of three major kinds of filaments—f-actin, microtubules and intermediate filaments—that form a cross-linked dynamic meshwork in the cytoplasm, providing shape and structure to the cell [1, 3]. The most dynamic constituent of the cytoskeleton, which is especially important in force generation and motility, is the filamentous actin (f-actin) network . The f-actin network is directly involved in the formation of lamellipodia and filopodia through polymerization of f-actin against the cell membrane . A third kind of cellular protrusions, blebs, are a result of the cortical actin network detaching from the cell membrane , and the convex shapes of adherent cells have been shown to result from myosin-II driven actin contractility .
For this study DUNN and DLM8 osteosarcoma cancer cells were used. The DLM8 line is derived from the DUNN cell line with selection for metastasis . Therefore, DLM8 is closely related to DUNN except for degree of its invasiveness. Both cell lines were a gift from Dr. Douglas Thamm (Colorado State University, CO, USA). They were cultured on glass detergent washed and air dried (GDA) substrates with standard culture conditions of 37°C and 5% carbon dioxide concentration in Dulbecco’s Modified Eagle Medium (DMEM) (Sigma Alrdich) in triplicates on the same day. DMEM was supplemented with 10% EquaFETAL Fetal Bovine Serum (Atlas Biologicals) and 100 Units/ml penicillin with 100 μg/ml streptomycin (Fisher Scientific-Hyclone). After 45 hours of culture, cells were incubated with different cytoskeletal drugs with description, conditions, and vendors listed in the Table A in S1 File for 3 hours. Then the cells were washed and fixed with 4% paraformaldehyde. Finally, they were fluorescently stained for nuclei (DAPI from Molecular probes) and actin (Acti-stain 488 phalloidin from Cytoskeleton, Inc). All the drugs were dissolved in Dimethyl sulfoxide (DMSO) (Fisher BioReagents) and to drop its effect on cell shape and actin structure control study was also treated with DMSO with the same molarity as other drugs. Then the cells were imaged using fluorescent microscopy. Representative images of each cell line treated with these drugs are shown in Fig 1.
In order to strike a balance between the throughput and the accuracy of the image processing, the image processing is fully supervised by the operator to reduce the number of artifacts and also automated as much as possible to speed-up the processing of the images. The image processing code is stream-lined into a consecutive step-by-step workflow which consists of four steps as follows. 1) A graphic user interface (GUI) enables the binary thresholding of the actin and nucleus images. While a thresholding value is suggested automatically by Otsu’s method, users can easily adjust the thresholding value using a slide bar in the GUI by visually checking the original image and the thresholded image displayed side-by-side in the GUI. 2) Cell declustering in the thresholded images is done using an optimized template of the open-source software CellProfiler . 3) The outputs from the CellProfiler are then visually examined by the operator and corrections can be made if necessary using the modules functionalized into this step. To facilitate the following analyses, each cell is centered and saved separately into one 1024×1024 image. Except for CellProfiler, all the other image processing codes were programmed in-house using Matlab (Mathworks) and are available on https://github.com/Wenlong-Xu/Image_Processing_Cell_Shape. Also available are the detailed protocol on how to configure the image processing codes and the CellProfiler template used for cell declustering.
In this paper we introduce and provide the TISMorph package to quantify cell shape and cytoskeletal structure based on two dimensional images of cell morphology and actin structure. The Matlab toolbox used to process the images and quantify the shape and structure of a cell is shared in GitHub repository to be used by others. These toolboxes can be found in the following addresses, https://github.com/Wenlong-Xu/Image_Processing_Cell_Shape, https://github.com/Elaheh-Alizadeh/Quantifiction-of-shape-and-structure. Some of the textural and morphological features calculated by TISMorph are similar to the measures calculated by CellProfiler(CP) . Geometric features of nuclei and cell, and gray scale measures calculated in this paper overlap with the measures calculated in CellProfiler by MeasureObjectSizeShape and MeasureTexture module. CP also calculates Zernike moments, though only to order 10, while we use Zernike moments up to order 30, because we find that fewer orders do not resolve objects sufficiently. Some of the measures we calculate are not implemented in the version of CP current during the time of submission. These are Fractal dimension, hull geometric measures, band based measures, and irregularity measures. However CP does calculate a few additional measures in the MeasureObjectIntensity module, which are statistical measures of intensity of objects such as mean and standard deviation of intensity, that are currently not included in TISMorph. Thus TISMorph includes significant additions to the previous state of the art as represented by the quantitative metrics calculated by CP.