4. Quantifying cellular phenotypes

Cellular phenotypes, such as cell morphology, the intensity or sub-cellular spatial distribution of labeled protein markers, are important biological properties of individual cells. This section shows how to extract quantitative measurements of cellular phenotypes from raw image data. A typical cellXpress workflow to quantify cellular phenotypes involves the following:

  1. Image background subtraction
  2. Quick plate analysis: quality control of the acquired images
  3. Cell segmentation
  4. High-content feature extraction
  5. Multi-dimensional profile-building using R

4.1. Background subtraction

Raw fluorescence images often have non-uniformly illuminated background. To minimize the effect of this problem on any downstream analyses, background subtraction should be carried out on the raw images. The cellXpress performs background subtraction using the rolling-ball algorithm implemented in ImageJ. This algorithm has only one parameter, the size of rolling ball, which should be set to at least the size of the largest object that is not part of the background. An ImageJ macro script is provided that performs background subtraction for all images within a plate directory. Instructions for installing and executing the script are given below:

Other software required:

  1. Java Development Kit (JDK) version: 1.6.24
  2. ImageJ version: > 1.47j
  3. PNGJ version: 0.96

Install JDK and ImageJ:

  • The simplest way to do this is to install ImageJ bundled with the JDK from here,
  • Then, start ImageJ and select Help > Update ImageJ….
  • All the other components needed can be found under "{cellXpress directory}/scripts/imagej". Please always use files from the latest release of cellXpress.

Compile the cX_SavePNG16.java plugin:

  • Copy the file "cX_SavePNG16.java" from "{cellXpress directory}/scripts/imagej" to "{ImageJ directory}/plugins".
  • Copy the file "pngj.jar" from "{cellXpress directory}/scripts/imagej" to "{ImageJ directory}/jre/lib/ext".
  • Start ImageJ, select Plugins > Compile and Run…
  • Select the "cX_SavePNG16.java" file which you saved in the ImageJ plugin directory above. The error message can be ignored.
  • Quit and re-start ImageJ. A new cX SavePNG16 option will appear under the Plugins menu. This installation only needs to be done once.

Run the background subtraction script:

  • First create the destination directory where the background-subtracted images will be saved. This will typically be a sub-directory under the Data root directory.
  • Start ImageJ and select Plugins > Macros > Run…, then select the background_subtraction_tif.java script to run.
  • Select the source directory that contains the raw images.
  • Select the destination directory you just created.
  • Enter the rolling-ball size parameter.
  • The resulting background-subtracted images will be saved to the destination directory.
images/Ch4_open-ImageJ.png
images/Ch4_imagej-marco-background-subtraction.png
images/Ch4_imagej-bgsubtract-choose-source-directory.png
images/Ch4_imagej-bgsubtract-choose-destin-directory.png

4.2. Quality control using Quick Plate Analysis

The Quick Plate Analysis tool facilitates quality control of the raw images acquired. This tool calculates and displays well-to-well variation in fluorescence intensities for each plate, allowing you to rapidly identify plates or wells containing poor-quality images, and re-acquire the data if necessary. Quick Plate Analysis can be carried out as follows:

  • [cellXplorer]: Analysis>Quick plate analysis
images/Ch4_quick-plate-analysis-open.png
images/Ch4_quick-plate-analysis-plot.png

4.3. Cell segmentation

To quantify cellular phenotypes at the single-cell level, individual cells in an image must first be identified and segmented into subcellular regions. The cellXpress uses a marker-based watershed algorithm for cell segmentation, described below. This algorithm requires at least two images per well position, one of which must be a DNA or nuclear-stain image.

  1. The image from the nuclear-stain channel is first segmented by applying a thresholding method to identify the DNA regions, followed by a watershed algorithm to split the connected cell regions.
  2. Next, a composite cell image is obtained by linearly combining the images from the remaining non-DNA channels. The composite image is then segmented to identify the regions stained with fluorescent markers, thus defining the whole cell region.

To perform segmentation

  • [cellXplorer]: Profiling>Segment cells for the selected wells
images/Ch4_segmentation-run.png

4.4. Object and region detection

Measurements of marker features in defined cellular sub-regions can provide information about the spatial distribution of markers within cells. The cellXpress provides an object detection module for detecting marker objects in eight different cellular sub-regions:

  • Whole cell region
  • DNA region
  • DNA boundary
  • DNA inner region
  • non-DNA region
  • non-DNA/cell boundary
  • non-DNA/cell inner region
  • non-DNA/cell peri-DNA region

Object detection is done using an adaptive thresholding algorithm, with parameters that can be adjusted:

  • [cellXplorer]: Markerset>Setting>Object detection
  • [cellXplorer]: Markerset>Setting>Region detection
  • [cellXplorer]: Markerset>Setting>Feature and region types
images/Ch4_set-object-detection.png
images/Ch4_set-object-detection-parameters.png
images/Ch4_set-region-detection-parameters.png
images/Ch4_object-detection-types.png

4.5. Feature extraction

The cellXpress offers four different types of commonly-used high-content features:

  • Cell morphology features
  • Fluorescence intensity features
  • Texture features
  • Moment features

These features can be calculated for individual cells over the eight different sub-cellular regions defined during segmentation.

  • Whole cell region
  • DNA region
  • DNA boundary
  • DNA inner region
  • non-DNA region
  • non-DNA/cell boundary
  • non-DNA/cell inner region
  • non-DNA/cell peri-DNA region

To perform feature extraction:

  • Select the feature types to extract

    • [cellXplorer]: Markerset>Settings>Feature and region types
    • [Feature panel]: Using the checkboxes, select the required features and the sub-cellular regions for which to extract them.
  • Run feature extraction

    • [cellXplorer]: Profiling>Extract features for the selected wells
images/Ch2_setting-feature-types.png
images/Ch2_feature-type-selection.png
images/Ch4_extract-feature.png

4.6. Cell counting

The cell counting tool computes the total and mean number of cells identified from the images for each well and displays the result for all the wells in the plate graphically. This function is only available after cell segmentation and feature extraction have been performed.

  • [cellXplorer]: Analysis>Count cells
images/Ch2_cell-count.png

4.7. Analyze intensity

The intensity analysis tool provides a quick graphical summary of the intensity features for the selected plate. This function is only available after cell segmentation and feature extraction have been performed. Using the Results menu, you can select a feature to display. For each feature, the mean values for the well are shown as a color map, with numerical values shown in the side-bar. An "X" in a well indicates that no data is available for that well. The intensity analysis results can be saved in graphical or compressed numerical data format.

  • [cellXplorer]: Analysis>Analyze intensity
  • [intensity analysis]: Use the Results menu to change the feature being viewed
  • [intensity analysis]: Use File>Save as to save the analysis results
images/Ch2_plot-intensity-distribution.png