EzColocalization: Versatile Quantitative Image Analysis

Inputs and alignment tab. (A). Inputs tab in the GUI. (B) General steps for the alignment of images. The cell identification image stack (phase contrast; left column), reporter 1 image stack (DAPI staining of DNA; center column), and reporter 2 image stack (Cy5; right column) are images of a previously reported bacterial strain (HL6320)^{15 (link)}. Scale bar is 2 μm. Reporters 1 and 2 images are pseudocolored. Red coloring in the second row of images indicates the objects identified by thresholding of the signal in each channel (“Default” algorithm in ImageJ). Following alignment of the images, pixels that overhang are removed and gaps are filled with pixels with zero value (yellow areas) so that all images have the same area in the common aligned region.

Cell identification and cell filters tab. (A) Cell Filters tab in the GUI. (B) Cell selection and watershed segmentation. Red coloring in the image in the second row indicates objects identified by thresholding of the signal in the cell identification channel (“Default” algorithm in ImageJ). Cells are the same as in Fig. 1. (C) Selection of cells based on physical features using the cell filters. Scale bar is 2 μm. Phase contrast image from Fig. 1. Red outline indicates the objects that were identified by thresholding (Panel B), and in the case of the right image, are within the parameter range(s) selected by the filter. (D) Selection of cells based on signal intensity using the cell filters. Phase contrast (cell identification image) and DAPI stain (reporter channel) images of bacteria (HL6187). Scale bar is 2 μm. Note: the lower of the two cells (no red border) has been removed from the analysis by the cell filter (that is, it no longer has the red cell outline).

Visualization tab. Data are from bacteria (HL6187) with labeled sodB::gfp RNA (Cy3 channel) and DNA (DAPI). (A) Visualization tab in the GUI. (B) Heat maps of Cy3 and DAPI signals for bacteria with “cell scaling” (defined in main text). Scale bar is 2 μm. (C) Scatterplot of Cy3 and DAPI for the cell on the left and outlined in white in Fig. 3B. (D) Metric matrix for TOS (linear scaling) for the cell on the left and outlined in white in Fig. 3B. F_T is the top percentage of pixels in the channel; for example, if F_T for Cy3 is 80% then it refers to the 80% of pixels with the highest Cy3 signal. Black color on the left column and bottom row indicate that TOS values are not informative when one threshold is 100%; that is, the overlap of two reporters can only be 100% if 100% of pixels are selected for at least one channel.

Analysis tab. (A) Analysis tab in the GUI for selecting default metrics. Note: this example is for two reporter channels (see Fig. 8G for 3 reporter channels). (B) Analysis tab in the GUI for users to code custom metrics. The example code provided is for measuring colocalization by Pearson correlation coefficient. (C) Example of a data table showing metric values for Pearson correlation coefficient (PCC) and some of the parameter values for some of the cells in the analysis. Label = the image and unique cell number to identify individual cells; Area = area of each cell in pixels; and X = the average x-value of all pixels in a cell. Data is from the example used in Fig. 3. (D) Summary report (“Log”) of the results in Fig. 4C. (E) Histogram generated from the results in Fig. 4C. The height of each bin is the relative frequency. The Count is the number of cells. Mean is the mean value. StdDev is the standard deviation. Bins is the number of bins. Min and Max are the minimum and maximum values of the lowest and highest bin respectively (which are shown immediately under the histogram). Mode is the mode value. Bin Width is the width of each bin within the histogram.

Application 1: Cell selection using reporter images and physical parameters. Images are rat hippocampal neurons labelled with an F-actin probe and anti-tubulin antibody visualized by fluorescence microscopy (see main text). (A) Workflow of the analysis. (B) Cell identification using the F-actin reporter and filters to remove small non-cell objects (yellow arrow) based on their size (i.e. Area option from the cell filters). Large yellow box in left panel is a zoomed in view of the smaller yellow box. Red outline of the neuron indicates it has been identified as an object (i.e. a cell) for analysis. Scale bar is 100 μm. (C) Heat maps with cellular normalization showing localization regions of signal intensity for the cell shown in panel B. Scale bar is the same as panel B. (D) Scatterplot showing relationship between the signal intensity for two reporter channels for a random cell in the sample. Pixels with the highest intensity signal for each reporter channel have the lowest intensity signals for the other reporter, which indicates anticolocalization (blue circles). Green dash lines indicate thresholds selected by Costes’ method. (E) Metric matrix for the median TOS (linear) value for all cells in the sample (n = 20). Green box indicates the threshold combination where F-actin and tubulin have the highest intensity signal (top 10% of pixels for each channel); the median TOS value is −0.36.

Application 2: Image alignment. Images are S. cerevisiae with TEM1 translationally fused to GFP and DAPI staining visualized by DIC microscopy and fluorescence microscopy (see main text). (A) Workflow of the analysis. (B) Cell identification by hand-drawn ROIs on a DIC image and creation of a binary image mask. Red outline indicates the boundary of the hand-drawn ROI. Scale bar is 3.5 μm. (C) Alignment of the reporter images using the binary mask image. Arrows indicate areas of misalignment that are corrected. Red outline is the same as for Panel B.

Application 3: Cell selection using signal intensity parameters. Images are whole adult C. elegans with GFP expressed from the clec-60 promoter and mCherry expressed from the myo-2 promoter that are visualized by bright-field microscopy and fluorescence microscopy (see main text). (A) Workflow of the analysis. (B) Selection of C. elegans so that only those individuals with an average intensity for the reporter signal that is above a threshold level are included in analyses. Left image shows the ROI manager with a list of ROIs that were hand-drawn around each C. elegans. Right image shows the reporter channel images with red outlines indicating the boundaries of the ROIs. C. elegans below the threshold level were excluded (yellow arrow) from the analyses by using the cell filters for signal intensity. Scale bar is 250 μm.

Application 4: Measurement of colocalization for three reporter channels. Images are of human bone cancer cells (U2OS) labelled as described in the main text. (A) Workflow of the analysis. (B) Images of cells in the cell identification and reporter channels. Top row are raw images. Bottom row, left image is the cell identification with pseudocolor (blue is the signal from Hoechst 33342 signal and green is the signal from phalloidin/Alexa Fluor 568 conjugate and wheat germ agglutinin/Alexa Fluor 555 conjugate) and boundaries of the ROIs in white (see main text). Bottom row (except left image) are heat maps for each of the three reporters with the boundaries of the ROIs shown. Signal intensity is indicated by the bar below each reporter image. Scale bar is 20 μm. (C) A three channel scatterplot for a single cell is shown for illustrative purposes only. (D–F) Metric matrices of median values for ICQ (D) TOS (E) and Manders’ colocalization coefficients M1, M2 and M3 (F) for all cells in the analysis (n = 66). Note: black color on metric matrix for ICQ indicates there were no pixels above all three thresholds for some cells, and therefore ICQ could not be calculated. (G) Analysis Metrics subtab for the Analysis tab for three reporter channels.