klt feature-tracking algorithm  (MathWorks Inc)

Journal: bioRxiv

Article Title: Assembly of a persistent apical actin network by the formin Frl/Fmnl tunes epithelial cell deformability

doi: 10.1101/680033

Figure Lengend Snippet: (a) Description of the method used to quantify to pulsatile Rho1 activity without cell segmentation (see Methods for more details and Supplementary Movie 16). Basically, isolated clusters of AniRBD::eGFP signal are detected using a DBSCAN algorithm. These clusters are then converted into surface ROIs using convex hulls and overlapping ROIs tracked over time to follow individual pulses. Filters such as min./max. area or min./max. duration are applied to reduce tracking mistakes. Finally, the AniRBD::eGFP pulse amplitude measurements are performed considering the maximum of total fluorescence intensity for each track. (b) MyoII pulses are automatically detected in time following the derivatives of high-pass filtered total medial MyoII levels (see Methods for more details). Pulse temporal landmarks have been defined as follow : ti : initial time; tdmax : max derivative; t max : max amplitude; td min : min derivative; t f : final time. (c) MyoII pulses are automatically detected in space by monitoring the centre of mass of the medial sqh::mKate2 signal over time (see Methods for more details and Supplementary Movie 22). The actual pulse centre used for the KLT analysis is then defined by averaging the position of the recorded centre of mass between td max and td min of the pulse. This time interval corresponds to the period during which the apical surface contract during a pulse. (d) Line plots : Averaged speed towards the pulse according to pulse temporal landmarks (see above) for different distance bins (see legend). The black arrows show the time of maximum speed. (e) Schematics of the numerical model. The black circle depicts the actomyosin pulse, and arrows depict forces applied to boundary elements. Only a fraction of boundary elements is represented.

Article Snippet: We first tracked these apical F-Actin structures using a Kanade-Lucas-Tomasi features tracking algorithm (KLT) , implemented in C (S.Birchfield, retrieved from https://cecas.clemson.edu/~stb/klt ) whose tracking results have been exported to Matlab for further processing.

Techniques: Activity Assay, Isolation, Fluorescence

Journal: bioRxiv

Article Title: Assembly of a persistent apical actin network by the formin Frl/Fmnl tunes epithelial cell deformability

doi: 10.1101/680033

Figure Lengend Snippet: (a) Live F-Actin (eGFP::UtrCH) localization in amnioserosa cells during DC comparing the dynamic of a pulsatile event in a WT and a frl 59/59 cell. Time series : images are extracted from the Supplementary Movie 21 at the indicated time-points, max-proj. (4 x 0.33 µm). The yellow frames show the pulse localization, the red and green outlines/arrows show, respectively, the contracting and the expanding parts of the cell. (b) Zoomed view on a contracting frl 59/59 cell displaying radial F-Actin filaments emanating from the pulse core and capturing the AJs. (c) Schematic representation of a contracting cell in WT or frl 59/59 condition. (d) Measuring the propagation of pulsatile contractility by following discrete apical F-Actin structures using a KLT tracking procedure. Left image : concentric circles showing the distance from a pulse (yellow cross). Right image : single time-point extracted from the Supplementary Movie 22, displaying the color-coded speed of KLT tracked structures toward the pulse centre. The white lines show the segmented cell boundaries. (e) Box plots : averaged speed toward the pulse centre of KLT tracked structures as a function to the distance to the pulse in WT vs frl 59/59 (left) or WT vs Frl OE (right) amnioserosa cells. Data are binned as indicated (distance ± 2.5 µm, e.g. the 5 µm bin contains all tracks within a 2.5 to 7.5 µm distance to the pulse). Line plots : speed toward the pulse as a function to the distance to the pulse averaged per bin. Each shade of grey in the background represents the typical size of an amnioserosa cell radius (~ 8 µm). (f) Representative simulations for different values of λ (Supplementary Movie 23). Left panels depict the initial condition, right panels depict the cell state upon maximal contraction. Green segments indicate that a boundary element is connected to the pulse. The pulse position is the same in the three examples. (g) Diagram : Measuring cell shape irregularity by comparing the convex hull (connecting vertices) and the segmented apical cell surface. Box plots : ratio between the surface of the inward + outward regions and the surface occupied by the convex hull (see ) upon maximal deformation. 200 iterations were performed for each value of λ . For each iteration, the pulse position is chosen randomly within the cell. (h) Line plots : Averaged speed towards the pulse vs. distance to the pulse during the contraction phase. Averages were performed from 200 iterations for each value of λ . For each iteration, the pulse position is chosen randomly within the cell. Box plots (e,g) : extend from 1 st (Q1) to 3 rd (Q3) quartile (Q3-Q1 = IQR), whiskers : S.D. (e) or : Q1 or Q3 ± 1.5 x IQR, horizontal lines : medians, black squares : means. Statistical significance (e,g) : two-sample t-test, NS : p > 5E-2, * : p < 5E-2, ** : p < 5E-3, *** : p < 5E-4, **** : p< 5E-5.

Article Snippet: We first tracked these apical F-Actin structures using a Kanade-Lucas-Tomasi features tracking algorithm (KLT) , implemented in C (S.Birchfield, retrieved from https://cecas.clemson.edu/~stb/klt ) whose tracking results have been exported to Matlab for further processing.

Techniques: