Image Analysis

Colocalization

Objectives of the lab:

The goal this week is to introduce you to one method for quantifying the kind of visual data you might collect during the course of lab this semester:

1. Use image analysis software, ImageJ/FIJI, to quantify the distribution of colocalization of proteins in transgenic mammalian NIH3T3 cells
2. Identify the potential strengths and weaknesses of using this type of analysis in this system

In order to open your images and analyze colocalization you will need to download the following program: ImageJ/FIJI. (https://imagej.net/Fiji#Downloads)

*Please download the FIJI software (and not ImageJ) as it comes preloaded with plugins necessary for opening your images).

You will also need to download the test files from the BI227 google drive, ER-RFP test.czi and Emerald-ELP1 test.czi found on the course Google Drive.

What is colocalization?

The spatial distribution of molecules in cells has significant impact on their function. In modern biological research, fluorescent microscopy is frequently used to examine the spatial distributions of molecules in cells. Fluorescence colocalization analysis–quantitative analysis of the overlap in the spatial distribution (also referred to as "co-occurrence") of two molecules/probes–is a powerful tool for determining whether two molecules localize to the same structure(s) in cells.

Figure 1: Schematic depiction of the spatial overlap of two probes (molecules) that display (A) no colocalization, (B) partial colocalization or © complete colocalization.

Visual evaluation of colocalization

For decades the prevailing method for determining whether two molecules "colocalized" in an image relied on the visual evaluation of the distribution of fluorescently labeled molecules in images of cells. To accomplish this fluorescence images were digitally superimposed on top of one another and either visually inspected for regions of signal overlap or, in cases where tools for displaying multiple-channel fluorescence images were used, inspected for regions of "merge" color (when Red and Green images are overlaid, "merge" color is conventionally displayed as "yellow") suggestive of colocalization. However, numerous results can be ambiguous. The problem is that an intermediate color, indicating colocalization, is obtained only if the intensities of the two probes are similar.

Figure 2: Fluorescence images of transgenic HEK293T cells expressing (A) Emerald:ELP1-25 (golgi) and (B) ER-RFP (endoplasmic reticulum). Images shown in (A) and (B) were superimposed using FIJI/ImageJ to highlight regions of signal overlap ( C).

Today, the determining whether two molecules are "colocalized" in an image is most often determined through computational analysis of the overlapping (non-overlapping) signal distribution and intensity within an image. This is most commonlu done using image analysis software, such as the FIJI/ImageJ software you will be using today. This software package includes a pluggin called COLOC2 which is able to calculate the pixel intensity correlation for an image using multiple methods including: Pearson, Manders, Costes and Li. The COLOC2 pluggin also performs automatic thresholding of images, generates scatterplots and conducts statistical significance testing.

To get started:

In FIJI: Start by opening your image files the pixel intensity correlations for an image.

(For practice, you can open the test images used in class last week: ER-RFP test.czi and Emerals-ELP1 test.czi found on the course Google Drive.)

1. Make the RFP “Red”:

To pseudocolor your image you will need to select a new Lookup Table:

• On the menu bar select: “Image” –> “Lookup Tables” –> “Red”

2. Make the Emerald/GFP “Green”:

• On the menu bar select: “Image” –> “Lookup Tables” –> “Red”

**4. Merge the images: **

• On the menu bar select: Image–> Color–> Merge Channels
• Select ER-RFP test.czi from the dropdown for C1 (red).
• Select Emerald-ELP1 test.czi from the dropdown for C2 (green).
• Make sure that “Create Composite” is checked as well as “Keep Source Images”

Your Composite image should look like this:

5. Let’s crop the image:

To crop the image to the region you are interested in:

• zoom in on you cell of interest
• select the “square” tool (selected in the screenshot below)

• click on your image and drag to select your rectangular region of interest (yellow box)
• On the menu bar select: Image–>Crop (The program will crop the image to the selected region)

Your cropped image should look like something like this:

6. Adjust/optimize the brightness/contrast of each of your color channels

• On the menu bar select: Image–> Adjust–> Brightness/Contrast
• On the menu bar select: Image–> Color–> Channels Tool
  • Color: turn on and off individual colors
  • Composite: combine colors (turn on and off individual colors)
• In the Channels Tool select "Color" from the dropdown menu and check "Channel 1"
  
  
  This will allow you to look at and manipulate just the "Red" or "ER-RFP" Channel in your image
• On the menu bar select: “Image” –> “Lookup Tables” –> “Fire”

This will change the Lookup Table in this channel to a "heatmap" where "black" indicates no signal and "white" indicates that a pixel has a signal intensity of 100%. You can use this information to adjust the brightness/contrast in your image to set the background to "0" or "black" and the brightest pixels in the image to "100%" or "white". This will help optimize the brightness/contrast settings in your image and remove background "noise" from your image

• Adjust the brightness/contrast in your image using the Brightness/Contrast tool
• On the menu bar select: “Image” –> “Lookup Tables” –> “Red”
• Repeat the above process for "Channel 2"

7. Change your cropped image to 8-bit

• On the menu bar select: Image–> Type–> 8-bit

8. Let's remove the background in your image

• On the menu bar select: Analyze–> Set Measurements and make sure that the only category of measurement selected is “Mean grey value”. Then select okay.

• select the “square” tool and draw a small square on a region of your your merged image that is outside but immediately adjacent to your cell (see example image below)
  
• On the menu bar select: Analyze–> Measure
  
• On the menu bar select: Process–> Math–> Subtract and enter the "Mean" (Mean grey value) you obtained in the previous step5 and select "OK"
  

**9. Split the cropped image into two separate channels **

Your Split cropped images should look like something like this:

10. If your cropped images still contain more then one cell (or fragments of cells) then you will need to set a region of interest (ROI) to analyze in the image. To do this:

• select either the “polygon" or "freehand" selections tool (the polygon selections tool is selected in the screenshot below)
  *
• Using the tool, trace your cell of interest
  

11. Launch the Coloc 2 plugin

• On the menu bar select: Analyze–> Colocalization Analysis–> Coloc 2
• Select the images according to which you want to be channel 1 (for the test I have selected EmeraldELP1-cell1.tif) and which to be channel 2 (ERRFP-cell1.tif).
• In the third drop down list selection, select the image/channel you want to use that has the correct ROI or mask image (for the test image, the ROI is set on the EmeraldELP1-cell1.tif which is in channel 1)
• You must now choose which “Algorithms” are run and which statistics you wish to calculate. Select: Manders' Correlation and "Costes significance Test".
• To tell the program to save the PDF result file when the OK button is pressed make sure to select "Show Save PDF Dialog".

Your COLOC2 settings should look like something like this:

• Select "Ok" to run the colocalization analysis

Note: Numerical results and image names are dumped into the ImageJ Log window as comma separated values, so you can copy paste, or save the log window contents, to then import them into whatever statistical package or spreadsheet in which you wish to analyze the results.

• Save your Colocalization results and then a window will appear that you can look over.

12. Looking at your results
The COLOC2 pluggin reports a great deal of information about the analysis of your cells.
For our purposes, please focus on the following:

• Slope: Slope of the line represents the "red to green" ratio, as a measure of both image intensity and colocalization. Ideally, the slope should be equal to 1 (y=x), however, it is more likely that one of the immunostained colors will be darker than the other causing the slope to tend more towards that axis. Good colocalization will give a scatterplot which is best fit by a linear curve, where the slope of this curve is representative of the ratio of immunostained colors.

• The Pearson correlation coefficient (PCC) and the Mander's overlap coefficient (MOC) are used to quantify the degree of colocalization between fluorophores.

  • The Pearson correlation coefficient (PCC) is derrived from the equation below:
  

The range for Pearson’s R value is ‐1 to 1 with ‐1 being total negative correlation, 0 being random correlation, and 1 being total positive correlation.

Note: Pearson’s is extremely sensitive to diferences in signal intensity between two images.

• • Mander's overlap coefficient (MOC) is derrived from the equation below:

Manders' expresses the proportion of each signal that overlaps with the other, thus, Manders' Overlap coefficients range between 0 and 1.

• Manders’ Colocalization coefficients M1 and M2:
  M1 is the percentage of pixels in Channel 1 that overlap with pixels in Channel 2 and M2 is the percentage of pixels in Channel 2 that overlap with pixels in Channel 1.

Note: M1 and M2 are sensitive to background but not sensitive to overlapping pixel intensities.

So…
The COLOC2 pluggin determines a background threshold for both images, removes this background from the images and reports:

• Manders’ Colocalization coefficients tM1 and tM2:
  tM1 is the percentage of pixels above the thresholded background in Channel 1 that overlap with pixels above the thresholded background in Channel 2 and M2 is the percentage of pixels above the thresholded background in Channel 2 that overlap with pixels above the thresholded background in Channel 1.

Note: M1 and tM1 and M2 and tM2 should not be equal.

For our analysis we will focus on the Manders’ Colocalization coefficients tM1 and tM2

• Statistics: Costes P-Value
  The Costes statistical significance test is used to determine with what degree of certainty you can be assured that colocalization exists in your images. A p‐value of 1 indicating >95% certainty that colocalization exists.
  The Costes test creates a scatterplot from pixel intensities in the two images then calculates an orthogonal line in the scatterplot with a slope determined by a least‐squares fit based on orthogonal regression. (This is a standard statistical procedure that I will not cover here.) Next, starting at the brightest point on that line T, it calculates Pearson’s coefficient for all pixels with values below T. Then T is sequentially lowered along the line and Pearson’s is recalculated until Pearson’s equals 0 (i.e. random correlation). At this point on the orthogonal line the threshold has been reached for each image. It is found by looking up the respective coordinate values on the x and y axes. After setting the threshold for both images, M1 and M2 are calculated.