#flow cytometry data analysis

Now you think that you know the pattern of expression of the molecule you are investigating. You think you know the sample you are investigating. You think you know how it should look when you stain your cell…Well, stop thinking so much and look at the data. Seriously. Many factors can affect the way your cells will appear in the cytometer, some of which you simply can't control. One such factor is the autofluorescence of your cells. If you do not properly display your cells, you can end up gating inaccurately. And you may be as surprised as I was when I bought this apple! Looking at only one parameter without going deep into the plots can mask the nature of your cells. Apple…orangetini anyone? (picture seen in CareerJoy blog, http://www.careerjoy.com) Autofluorescence in biological samples is intrinsic to some cellular components, and it varies from cell to cell. It is dominated by reduced pyridine nucleotides (NAD(P)H) and oxidized flavins (FAD), although other components of the cell may account for their emitted fluorescence upon laser excitation (there are even more in plants!). There may be ways to reduce the impact of autofluorescence (wait for a post on this!), but you can certainly gate around it when you have it. Autofluorescence is most evident in the FITC (FL1) and PE (FL2) channels. In this example you can see how showing a histogram does not fully reflect the origin of the fluorescence and the distribution of the populations. Cells expressing GFP which fluoresce in FL-1 By making a diagonal gate around the GFP positive population in the dot plot (rather than the histogram) it is possible to more accurately define the population. If we analyze the same plot, we can see the origin of the fluorescence for different cells: The origin of the fluorescence for the cells in the different circles are: Red, autofluorescent (showing in both FL1 and FL2) Black, FL1 emission comes from both GFP specific and autofluorescence, with little or no emission in FL2 Purple, GFP specific emission only in FL1 and high autofluorescence in FL2 In this situation is always useful to have an unstained control, or an FMO (Fluorescence Minus One) control. This will help to set up your gates. So next time someone tells you something like, YES I SWEAR, THAT POPULATION JUST EXPLODES WHEN I TREAT MY MICE WITH SUGAR-SWEETENED PBS...ask for the data first. For further reading, we can recommend you this paper: J Immunol Methods. 2012 Dec 14;386(1-2):101-7. It is regarding the controversial description of FoxP3-expressing macrophages; the expression could be autofluorescence. If you have any questions or comments, please just email us, [email protected]

OK, so once you have impeccably processed your sample, given them a hand-made, unique staining and ran them in your cytometer without the machine clogging or the computer freezing (I bet you've found that FACSDiva™ is quite temperamental, to honor its name…), then is time to analyze your samples. Easy peasy right? Not quite. Depending on the complexity of your experiment, it can take you as much time to analyze it as took to actually do it on the bench. At this point, if you are not very experienced in flow cytometry data analysis, or if you are doing the experiment for the first time, then you will have to play around with how to set your gates, what kind of plot to use and decide at the end what is the best way to present your data. Just when you thought you were done with the experiment…you have to analyze it. There are several kinds of dot plots. Many researchers have a fondness for the so called contour plots. Their advantage shows for samples where only a small number of events is collected. Contour plots display the relative frequency of the populations, regardless of the number of events collected. The plots show an enriched dendritic cell (DC) population from mouse spleen on which only a few hundred events could be collected. To the left is a regular dot plot showing all events. To the right is a contour plot with 20 contour lines. The second plot in this case displays better the DC population of interest. However, contour plots are not good in the opposite situation. In the following case, the same DC population is being studied, but in the context of its natural occurrence in the spleen, here it is not enriched. If we keep the previous settings  for the contour plots, the population is barely visible. This is because a lot of total events were collected, and the only previous gate is viability. DCs represent only a small fraction of the spleen, so if we gated on this plot we will be excluding a lot of DCs, as they are not optimally displayed. A possible solution is to show the outliers, as in the middle plot, or use a different plot, such as the pseudocolor, or a regular dot plot. Another kind of widely used graphic is the histogram. A histogram, as opposed to a bivariate graph, displays only one parameter and the number of cells that show a given fluorescence for that parameter. Taking the same example, if you have little experience working with DCs, and were going to define them only based on expression of CD11c, you may end up setting an inaccurate gate. Overestimation of cells due to inaccurate gating. The histogram to the left shows a generous gate of positive events, representing 4.5% of total splenocytes. However, when an MHC II vs CD11c plot is used, a more accurate gating shows that DCs are only 1.2% of the total population. The black gates show the cells that are not DCs, but could be taken as such with suboptimal gating. So, choosing the best way to display your data will help you convince your peers of what you are proposing. You'll avoid, more or less, explicit expressions of disbelief in public regarding the way to identify your cells. By the way, did you notice all my examples are based on DCs? Big time fan, DCs are cool! Don't agree? Dare to let me know:[email protected]. Anyway, stay tuned for more tips on gating and how histograms can further mislead your analysis! Wait, you are looking at T regs? With those gates? Really?