Fig 1.
Overview of Object-Oriented Polarization Software (OOPS).
(A) Hierarchical data structure used by the software. Projects are organized into “groups” representing biological replicates or experimental conditions, each containing a certain number of “images”. Each image contains “objects”, which store properties and statistics calculated for individual structures in the image. (B) Simplified processing pipeline for a single “image”, which includes flat-field correction, segmentation, calculation of FPM statistics, and object feature extraction. (C) Screenshots from the OOPS GUI showing examples of interactive data processing including examination of data, design of custom segmentation schemes, adjustment of image masks, and object manipulation. (D) Screenshots showing examples of the different image, plot, and data table visualizations available in the software.
Fig 2.
Object-based analysis reveals broad biological diversity in desmosomal cadherin architecture.
Desmosomal cadherin order probes expressed in A-431 cells, imaged with FPM, and analyzed with OOPS to demonstrate object-oriented FPM image analysis. (A) Representative intensity (top left) and order (p, top right) images of the Desmocollin 2b extracellular order probe (ECTOb). A single cell-cell border containing several desmosomes is indicated with a rectangular ROI and magnified below. Objects detected after segmentation are enclosed by white boundaries. Two objects that differ in order and signal-to-background ratio (S/B) are indicated by arrows. (B) Closer inspection of the low S/B object in (A). (Top left) Object intensity images at each excitation polarization, normalized to the maximum intensity in the stack. Arrows indicate the direction of the excitation field. (Top right) Average intensity image, normalized to the maximum intensity. Labels indicate regions used to determine local S/B (S: signal; B: background). (Bottom left) Pixel intensity stacks normalized to the total intensity and fit to a generic sinusoid (gray: individual pixel fits; blue: individual pixel azimuths; black: average of all fits). (Bottom right) Object order image. Pixels used to calculate mean order are enclosed in a white boundary. (C) Same as (B), but for the high S/B object indicated in (A). (D) Swarm plots showing mean order () for each object in the desmosome dataset, grouped by construct: ECTOb (red), CYTO (blue), LINK (green), ECTOa (good) (purple), ECTOa (poor) (yellow). (E) Scatter plot of
versus local S/B for each object, grouped as in (D). (F) All objects across all groups were sorted into low and high S/B clusters using k-means clustering. Objects are represented by their average intensity images, which are tiled and stitched together within each cluster: Cluster 1 (light blue; low S/B) and Cluster 2 (dark blue; high S/B). (G) Same as in (D) but grouped by both construct and cluster. (H) Same as in (E) but grouped by cluster.
Fig 3.
Relative azimuth calculations reveal nanoscale cadherin geometry in the desmosome.
(A) Representative “S-shaped” object from the ECTOb dataset. (Top left) Object intensity images at each excitation polarization, normalized to the maximum intensity in the stack. Arrows indicate the direction of the excitation field. (Top right) Average intensity image, normalized to the maximum intensity. (Bottom left) Pixel intensity stacks normalized to the total intensity and fit to a generic sinusoid (gray: individual pixel fits; blue: individual pixel azimuths; black: average of all fits). (Bottom right) Object azimuth image. Pixel values represent the angle of the azimuths with respect to the excitation field in I0° (αimage). Background pixels are partially masked to highlight the object. (B) Average intensity image of the object in (A) with overlaid azimuth sticks, colored according to the direction, αimage. (C) Simplified overview of the midline detection algorithm showing—from left to right—the binary mask defining the object; coordinates of the 8-connected perimeter pixels of the mask; boundary coordinates after dilation, smoothing, and linear arc interpolation; Voronoi diagram of the adjusted boundary points; and final midline detected from the central most edges of the Voronoi diagram after smoothing and interpolation. (D) Same as in (B), but with the midline overlaid and azimuth sticks colored according to their direction relative to the nearest midline tangent (αmidline). (E) Polar histogram showing the distribution of for all objects in the ECTOb dataset. Object
values initially in the range [−π/2,π/2] are duplicated and shifted by π to show each pair of equivalent, opposite directions. (F) Same as in (E), but for object
directions.
Fig 4.
Quantifying cytoskeletal reorganization in cells grown under flow.
(A) Human umbilical vein endothelial cells (HUVECs) grown in static conditions, labelled with AF488-phalloidin, and imaged with FPM. From top to bottom: average intensity image, azimuth-order-intensity HSV image, and binary mask showing locations of detected filaments. The region indicated by a square ROI is shown magnified to the right. (B) Same as (A) but for HUVECs grown under fluidic shear stress (FSS). White arrow indicates the flow direction. (C) Polar histograms showing (left) and
(right) distributions for filaments in cells grown statically. (D) Same as (C), but for cells grown under FSS.
Fig 5.
Object-based analysis reveals hidden relationships between F-actin filament architecture and morphology.
(A) Swarm plots showing mean order () for each segmented filament in cells grown in static conditions (green) or under fluidic shear stress (FSS, yellow). (B) Swarm plots showing azimuth circular standard deviation (s0) for the same groups in (A). (C) Scatter plot matrix exported from OOPS showing relationship between filament length (L) and FPM order and azimuth statistics for all filaments across both growth conditions, sorted into two groups: “Short” (blue, L < 17.5 μm) and “Long” (orange, L ≥ 17.5 μm). (D) Magnified version of the scatterplot highlighted by a black square in (C), showing the relationship between L and
. A dashed black line denotes the cutoff point between “Short” and “Long” filaments, which was chosen to approximate the reported persistence length (Lp) of phalloidin-stabilized F-actin, Lp = 17.5 μm. (E–F) Same as in (A–B) but grouped based on growth condition and filament length.