Lesson 7: CFSE – A Detailed Example

Here is a more detailed look at the cfse analysis procedure discussed in lesson 7.

CFSE – Where to begin

A good first step prior to attempting a curve fit is to take a look at all of the data files together in order to determine which file to analyze. As mentioned in the lesson, the samples are counterstained with PE, so I will display FL1 (CFSE) versus FL2 (PE) as per the example. While it may not have seemed obvious looking at the files individually, the data files actually represent two sets of time courses. Files 005 – 008 represent one time course, while files 009-012 represent another.

Looking at the first series, notice how there is only one significant cfse peak in the image. As time moves on, more and more “peaks” start to appear. Notice also, that the newer peaks appear lower on the FL1 axis due to the fact that the CFSE stain contained in the parent cell is split between the daughter cells; in other words, the newest cells have the lowest CFSE staining. Notice also that the PE positive and PE negative populations have slightly different kinetics with different relative distributions between each division.

Curve Fitting

Let’s first take a look at the automated curve fit method in Weasel. For this example, I’ve chosen to look at file #11. As the instructions indicated PE counterstaining, I will first create a region around the PE-postitive events as in the image aboce. I will call my region PE Pos.

I will then convert my region into a gate using the create gate option.

The next step is to create a New Histogram. Notice how I have chosen only to show the FL1 data and that the PE Pos gate is selected.

Once the histogram is created, simply right-click on the plot and choose curve fit. This will bring up the Set Fit Parameters windows. Choose the CFSE Fit tab. I generally leave the default options in place unless I expereince problems with the fitting process.

In the histogram, grab the (red) left marker and move it to the beginning of the data peaks. More precisely, this should be set for the maximum point of autofluorescence (background), but as you are not provided with this data, simply moving the marker to the start of the CFSE peaks as above will suffice. Next, click Do CFSE Fit.

It will likely require several passes to successfully fir the CFSE data. If you are presented with a convergence message window such as this, simply click continue until the fit is complete.

After completion, you will be able to read out the number of measurable divisions (~8) and the percentages in each generation. Notice that the highest number of cells exist in generation six, 26%, according to the curve fit model.

 

Manual Histogram Fit

It is also possible to do a manual “fitting” of the data. Let’s return to our FL1 histogram gated on the PE Pos region. I can choose to simply draw regions (right-click, Create Region) over each of the peaks and show the resulting statistics. This is somewhat tricky on the histogram as generations 2 and 3 overlap quite a bit. In the example shown, I am reporting two statistics: %Total and %Gated. I am primarily interested in the %Gated, those cells that fall in the PE Pos region. Notice that the maximum number of cells falls within the W5 gate which represents the sixth generation. The reported percentage is ~24% which is close to that of the curve fit process, certainly close enough if we are just looking for number of generations and relative distributions.

 

 

 

 

 

 

Manual Dot Plot Fit

A final way of fitting the data would be to manually draw 2D regions on the FL1/FL2 dot plot itself. In the above example, I found it a bit easier on the eyes to create a second plot showing only the data in the PE Pos region. I then created a series of rectangular regions for each cellular generation. Notice that in this method, it is somewhat easier to distinguish the 2nd and 3rd (R2 and R3) generations. As in the other fits, R3 represents a very small number of cells. Again looking at the %Gated state, R6, the sixth generation, represents the greatest population of cells at ~25%.

As can be seen from the above, there are many ways to approach a CFSE fit. The method chosen depends upon the information needed. If quantitative data is necessary, a modeled fit probably represents the best choice. If, however, relative distribution information is all that is needed, a quick manual fit will provide equally valid data.

Again, special thanks to Sean Linkes for providing the CFSE data.