Fig 1.
Choice of ECM protein coating and experimental design.
(a) A heat map shows the values of Pearson’s correlation coefficient measuring the association between ECM protein concentration and the number of cells attached. (b) Substrate stiffness values investigated in this study. (c) In the designed experiment, MDA-MB-231 cells were seeded on silicone substrates of different stiffness coated with collagen type I. Afterwards, cells were cultured for 24 h, fixed, stained for biomarkers of interest and imaged using confocal microscopy.
Fig 2.
Generation of single-cell profiles and data analysis workflow.
(a) A flow chart illustrating the key steps of image-based cell profiling and data analysis workflow. (b) Schematic representation of the image segmentation process: cells were labelled with WGA and DAPI to visualise cytoplasm and nuclei. Extracted cell outlines were used to quantify the intensity of E-cadherin, vimentin or cytokeratins within each cell. (c) Illustrations of the measurements calculated for each cell: geometric parameters, intensity and texture of the fluorescent signal, and measurements of the context. (d) Summary of the final dataset. Left column indicates the proportion of cells captured at each stiffness level. Morphological and contextual features of the cells were calculated (left); additionally, intensity and texture of the fluorescent biomarker signals were measured (right). Each row of the heatmaps corresponds to a single cell and each column represents a single feature. Z-score normalisation was applied to each feature to allow for direct comparison regardless of the scale. Right column indicates z-score values.
Fig 3.
Distinct adaptations to substrate stiffness by TNBC cells.
(a) An increase in cell density compared to baseline (dashed line) after cells were cultured for 24 hours on substrates with different stiffness values, n = 4 replicates per stiffness value. (b) Analysis of the association between the top 10 parameters modulated by substrate stiffness. Most measurements are related and thus are highly correlated with each other. (c) Changes in the distributions of NCR, cell perimeter, area, smallest diameter, circularity, and distance to the nearest cell in response to substrate stiffness. Distributions are reported using box plots: a box shows the median value, first and third quartiles, whiskers indicate median +/-1.5 * IQR. Significance of the difference assessed by Welch’s t-test (‘***’: p < 0.001, ‘**’: p < 0.01, ‘*’: p < 0.05), n = number of cells, see Table A in S2 Text.
Fig 4.
Identification of cell morphs using machine learning.
(a) The results of HC applied to 826 cells based on similarities between single-cell profiles consisting of 150 morphological and contextual features. Left column indicates three cell morphs (green, blue, and red) identified in the total population of cells. Each row of the heatmap corresponds to a single cell and each column represents a single feature. Subgroups of features are depicted on the bottom. (b) Representative objects (medoids) of each of the three identified cell groups. (c) Top six features ranked by their contribution towards cluster separation. Distributions of the values are reported using box plots: a box shows the median value, first and third quartiles, whiskers indicate median +/-1.5 * IQR. Significance of the difference assessed by Welch’s t-test (‘***’: p < 0.001). (d) Projection of the data on the first two PCs and visualisation of the identified cell morphs. “×” markers indicate the centroids of the groups. (e) Proportion of cells of each morph across stiffness values. Bar plots represent average proportions, error bars indicate 95% confidence intervals, values calculated by grouping cells by images. The inset shows the total number of cells in each cluster.
Fig 5.
Representative images of cells cultured at different stiffness levels corresponding to the three cell morphs, as determined by the clustering algorithm. The bottom row provides examples of Clumped cells classified as one of the three cell morphs by the RF classifier. Cell membranes were visualised using WGA staining.
Fig 6.
Changes in expression of biomarkers associated with epithelial and mesenchymal phenotypes.
(a-c) Mean fluorescent signal from E-cadherin, vimentin, and cytokeratins, respectively, across substrates of different stiffness. (d) The ratio of fluorescence intensity of cytokeratins to vimentin (CVR) calculated for all stiffness values. For (a-d) values are mean, error bars indicate standard deviation. Orange line shows the mean values and the band indicates 95% confidence intervals. Significance of the difference assessed by Welch’s t-test (‘***’: p < 0.001, ‘**’: p < 0.01, ‘*’: p < 0.05), n = number of cells, see Table A in S2 Text. (e) Box plots comparing mean fluorescent signal from E-cadherin, cytokeratins, and vimentin in cells belonging to three cell morphs. There appears to be no difference in CVR between the three cell morphs as can be seen in the rightmost plot. Distributions of the values are reported using box plots: a box shows the median value, first and third quartiles, whiskers indicate median +/-1.5 * IQR. Significance of the difference assessed by Welch’s t-test (‘***’: p < 0.001, ‘**’: p < 0.01, ‘*’: p < 0.05).
Fig 7.
Comparison of individual and clustered cells cultured at 64 kPa.
(a) Example images of individual and clumped cells cultured at 64 kPa. Cells stained with DAPI (blue) and WGA-FITC (green) to visualise cell nuclei and cytoplasm, and for E-cadherin (magenta). (b) Key differences between Single and Clumped cells. (c) Cell morph composition of Clumped cell population. (d) Mean intensity and mean edge intensity of E-cadherin in Single and Clumped cells. Schematic illustrations of the measurements are provided on the bottom. (e) Fraction of the total cell fluorescent signal located along the edge of a cell. Single and Clumped cells at 64 kPa are plotted separately. (f) TNBC cells (purple) displayed dramatically different morphological and colonisation patterns in the parenchymal (left) and stromal (right) compartments of our 3D liver tissue model. For (b), (d) and (e), values are mean, error bars indicate standard deviation. Significance of the difference assessed by Welch’s t-test (‘***’: p < 0.001, ‘**’: p < 0.01, ‘*’: p < 0.05).