Compensation HOWTO

Last modified Mon Sep 29 11:50:38 PDT 2003 - jdc


This document aims to provide an insight into flow cytometric compensation and to be a simple guide to compensating your samples.


Theory.

Fluorescence.

A fluorescent compound (fluorophore) absorbs incident photons and emits photons of lower energy. Lower energy photons have a longer wavelength, thus for example fluorescein absorbs blue light and emits green, phycoerythrin absorbs blue and green and emits yellowish-orange. Every fluorophore has a characteristic emission spectrum (and a characteristic excitation spectrum as well, but that is beyond the scope of this discussion). Most emission spectra exhibit a tail extending toward longer wavelengths; this is a consequence of the physics of fluorescence. Figure 1 shows the emisison spectrum of FITC:

Fig. 1
Fig. 1

How the flow cytometer measures fluorescence.

In order to separate fluorescence emission from the excitation light source and to resolve different colors, flow cytometers typically use a number of bandpass and dichroic filters. Simplistically, we can consider each detection channel to have one bandpass filter. The first fluorescence detector (frequently called FL1) is often equipped to detect green fluorescence with a medium bandwidth green filter. In the Becton-Dickinson FACScan a 530/30 filter is used. This filter can be considered to pass light from 515 nm to 545 nm and is denoted by the shaded region in figure 2:

Fig. 2
Fig. 2

Total light collected by the FL1 detector is denoted by the green shaded area.

Spectral overlap.

If we add a typical phycoerythrin bandpass filter (FL2, 585/42 in the FACScan) to the above diagram, we can see that this filter is able to collect some light emitted by FITC because of the tail in the FITC emission spectrum. Similarly, even a red-collecting channel (FL3, 650LP in the FACScan) can detect a little FITC fluorescence.

Fig. 3
Fig. 3

The need for compensation.

Figure 4 below extends on the previous diagrams by adding a line representing the emission spectrum of phycoerythrin (PE) (red line). It is now apparent that not only does some of the FITC fluorescence appear in the FL2 detector (region A), but a little of the PE fluorescence also appears in the FL1 detector (region B).

Fig. 4
Fig. 4

In order to properly analyze multicolor flow cytometry experiments it is necessary to employ a mechanism called color compensation. Specialized circuitry in the flow cytometer is used to subtract a portion of one detector's signal from another, leaving only the desired signal. In the above example, region A represents unwanted FITC fluorescence appearing in the FL2 detector. The cytometer can be programmed to subtract from FL2 a percentage of the signal present in FL1. This percentage depends on the shape of the spectrum and the characteristics of the bandpass filters but not on the intensity of the fluorescent signal. The same percentage may be therefore be subtracted regardless of whether the signal is strong or weak.

Compensation in practice.

Single color compensation.

The worked example below is derived from murine lymphocytes stained with FITC-anti-CD8 and PE-anti-CD4. Samples were analyzed with a Becton-Dickinson FACScan. Gating was performed using characteristic forward versus orthogonal light scatter (FSC vs SSC). Dot plots were constructed for FL1 (green fluorescence) on the X-axis versus FL2 (yellow/orange fluorescence ) on the Y-axis. First, a look at cells stained with only FITC-anti-CD8:

Uncompensated Compensated  
Uncompensated Compensated  

No compensation was applied to data collected for the left hand panel and the bright FITC positive cells show an appreciable signal in the FL2 detector. Compensation (FL2-%FL1) was applied and the results are shown in the right hand panel. For correct compensation, the median FL2 fluorescence intensities must be identical for FL1-dim and FL1-bright cell populations. Uncertainties in measurement of fluorescence intensity contribute to the higher FL2 coefficient of variation (c.v., "spread") observed for the FL1-bright population.

The procedure to correctly set compensation in the above example was as follows: Voltages on the FL1 and FL2 photomultipliers were set such that the dim ("negative") cells produced a population well clear of both axes and the bright FITC-positive cells were on scale. The FL2-%FL1 compensation control was then gradually increased until the median FL2 channel number of the FITC-bright cells was the same as the median FL2 channel number of the FITC-dim cells. Note that it is important to start with the negative cells well clear of the X-axis because the higher c.v. noted above may otherwise cause the compensated population to apparently extend below the axis. Since it is not possible for data points to have negative values in this system, such points are set to channel 0; such loss of data would make it impossible to correctly compute the median fluorescence.

The figure below illustrates the compensation procedure. Notice that the FITC-bright cells are moved downwards:

Compensation
Compensation procedure

Dual color compensation.

The example below extends the above to two colors. In the left hand panel, uncompensated data are shown for a mixture of unstained, FITC-anti-CD8 positive and PE-anti-CD4 positive cells. Notice that the CD8-bright cells which should ideally only show a signal in the FL1 detector also show a large signal in the FL2 detector. Conversely the CD4-bright cells which should ideally only show a signal in the FL2 detector also show a small signal in the FL1 detector. Compensated data are shown in the right hand panel. The compensation procedure involved setting FL2-%FL1 to approximately 19% and FL1-%FL2 to approximately 0.6%, and is illustrated in the lower panel.

Uncompensated Compensated  
Uncompensated Compensated  
Fig. 3
Compensation procedure

The two dot plots below illustrate the effect of insufficient FL2-%FL1 compensation (undercompensated) or too much (overcompensated). In the undercompensated case, the FITC-bright cells still show appreciable signal in the FL2 detector which on subsequent analysis might be misinterpreted as weak PE positivity. In the overcompensated case, a portion of the FITC-bright cell population has been forced into FL-2 channel 0; not only does this make it impossible to correctly assess the FL2 median, it also would cause underestimation of FL2 brightness of double-positive cells in subsequent samples.

Undercompensated Overcompensated  
Undercompensated Overcompensated  

Multiple color compensation.

Armed with knowledge of the emission spectra of the fluorophores involved, the above strategy may be used to apply compensation to any set of colors desired, subject to hardware and software limitations. Traditionally, electronic hardware compensation has been provided in flow cytometers because it is applied to the analog signal before digitization and produces superior results when digitization is limited to 10 bits (1024 channels) or less. A fully featured hardware compensation matrix may not be available; inter-beam compensation is particularly difficult from an engineering point of view and many multi-laser cytometers do not offer it. In the event that hardware compensation is not available for the desired pair of parameters, the investigator must use software which allows it (e.g. FCS Express, FCS Press, FlowJo, WinList (There may be others - tell me!). For those whose cytometers digitize to 16 bits or thereabouts, software compensation is probably preferable because the original experimental data are not modified in any way.

Two color compensation quick reference.

Quick reference protocol for 2 color compensation.

e.g. FITC or GFP (green) and PE (orange).

You will need the following samples:

Ensure that the green-only and orange-only cells are at least as bright, preferably brighter, than the brightest test samples you expect. You cannot correctly set up compensation with dimly stained cells!

Ensure you have washed the positive cells twice to remove residual fluorescent antibody.

Protocol.

Typical values for the FITC-PE compensation matrix on a FACScan instrument with FL1 and FL2 at identical voltages and between 400 and 650 volts are FL1-%FL2 = 0.4% and FL2-%FL1 = 19.6%

Graphical summary.

Uncompensated Compensated  
Uncompensated Compensated  

Three color compensation quick reference.

Setting up 3 color compensation is slighty more complex than 2 color compensation but is basically the same procedure repeated twice for different color combinations.

Quick reference protocol for 3 color compensation.

e.g. FITC or GFP (green), PE (orange) and Cychrome (red).

You will need the following samples:

Ensure that the green-only and orange-only and red-only cells are at least as bright, preferably brighter, than the brightest test samples you expect. You cannot correctly set up compensation with dimly stained cells!

Ensure you have washed the positive cells twice to remove residual fluorescent antibody.

Protocol.