(Go back one page to generalized compensation, or back to the introduction)

Autofluorescence throws a small kink into compensation (as it does into many things), but, as it turns out, does not change the ability to deconvolute spillovers. Cellular autofluorescence is present in all channels to varying extents, and provides a background (that varies from cell to cell). There are three ways to deal with autofluorescence.

One way is to devote a single channel to measure autofluorescence. Because the autofluorescence spectrum of cells is generally very similar, we can just treat autofluorescence as one more type of fluorescent molecule. Now, by compensation, we can actually correct for the contribution of autofluorescence to all channels. This process can significantly enhance sensitivity for detection of low-density antigens. Autofluorescence compensation is more fully described in published references (see Roederer & Murphy, "Cell-by-cell autofluorescence correction for low signal-to-noise systems: application to EGF endocytosis by 3T3 fibroblasts," 1986, Cytometry 7:558; and Albert, Parks, & Herzenberg, "A single laser method for subtraction of cell autofluorescence in flow cytometry," 1987, Cytometry 8:114).

The second way is to simply ignore it--apply the matrix algebra described previously still works to make the channels independent (i.e., with only autofluorescence and the specific fluorescent molecule of interest contributing to the channel). However, the resulting value is no longer a "true" fluorescence, but is, in general, a lower value than the "true" value. Nonetheless, the compensated fluorescence value is still proportional to the concentration of the fluor. What this means is that the compensated fluorescence value in any particular channel will not be equal to the value you would get by staining with only that fluorescent molecule and not performing compensation. Since we almost never have to perform this comparison, there is no problem.

Which leaves us the third way--an exact treatment of autofluorescence. This requires that we enter the realm of affine mathematics... this topic is dealt with in detail by Bagwell & Adams ("Fluorescence spectral overlap compensation for any number of flow cytometry parameters," 1993, in: Annals of the New York Academy of Sciences, 677:167-184).

Simple compensation in the presence of autofluorescence, therefore, works just fine: the resulting values are independent of the presence of other reagents, and are proportional to the amounts of the fluorescent molecules present. That's all we really need for flow cytometry.

Note, however, that if the autofluorescence of the stained cells in the compensation sample is different than that of the unstained cells in the compensation sample, then the computed compensation will be incorrect! Thus, you could not use FITC CD14 (staining highly autofluorescent monocytes) to compensate against unstained lymphocytes (which have low autofluorescence).

Go on to the requirements for proper compensation.