Quantitative imaging and analysis reveals that slow migrating cells increase in area while fast migrating cells decrease in area.

<p>(A) Live H1299 P/L cells expressing EGFP-paxillin and RubyRed-LifeAct were imaged and segmented in order to identify individual cells and their cohort of Cell-Matrix Adhesion Complexes (CMACs). EGFP-paxillin and RubyRed-LifeAct channels are displayed in inverted gray scale (high intensity is black). The segmentation image shows the EGFP-paxillin channel with the cell border identified in blue, CMAC borders in red and CMAC major axes in cyan. Scale bar: 10 μm. (B) By comparing consecutive frames, protrusions (green), retractions (red), short-lived (blue) and stable (gray) regions were identified, as described in Materials and Methods. White circles indicate the locations of CMACs. (C) The average size of each type of dynamic cell region (per frame) in the dataset was calculated and stratified per quintile of Cell Speed. The total height of each bar (the sum of protrusion, retraction and short-lived areas per frame) corresponds to the Dynamic Cell Area. We observed that this quantity increases with Cell Speed. (D) Scatter Plot: The net value of protrusion minus retraction areas (delta (Δ) Cell Area, μm²) is shown as a function of Cell Speed. The density of observations at a given Cell Speed (Cell Speed conditional density) is color-coded following log transformation, enabling better observation of trends in Δ Cell Area values given changing Cell Speed. A linear fit (Pearson’s correlation coefficient <i>r</i> = -0.27, <i>P</i> = 4.39·10⁻²³³) of the relationship is indicated (black line). Probability distributions: of Cell Speed (X axis), and; Δ Cell Area (Y axis, right) show a heavy-tailed distribution of Cell Speed where most cells are slow moving and few are fast moving, while Cell Area changes are approximately symmetrical overall. Notice, however, that the relatively few fast moving cells are decreasing in area substantially (retraction area is much larger than protrusion area), while the numerous slow moving cells only grow slightly, on average. (E) Mean Cell Speed autocorrelation coefficients (Y axis) are plotted conditioned upon the mean Cell Speed (X axis) of each cell. Autocorrelation values were calculated per cell trajectory with a maximum time lag of 1 h (12 frames). A linear fit (Pearson’s correlation coefficient <i>r</i> = -0.32, <i>P</i> = 0.00037) of the relationship indicates that autocorrelation of Cell Speed is lower in cells with higher average Cell Speeds. This indicates that the temporal persistence of Cell Speed correlates negatively to Cell Speed itself. See also <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135204#pone.0135204.s005" target="_blank">S1 Movie</a>, showing the same cell as in A-B.