cells were off scale (i.e., piled up in the first channel) and the CY was very high. At 450 Y, about half of the cells were on scale (left panel, dashed lines), but the CY was still relatively high. This would be the lowest recommended setting, and would be used if very bright events were to be acquired in FLI. The CY continued to decrease with increasing voltage and was close to its lower asymptote at 500 Y (right panel, dashed lines). This would be the preferred setting for the majority of cases. It is important to note that this setting is optimized for peripheral blood lymphocytes, which have relatively little autofluorescence. Different settings may be required for cultured cells or disaggregated solid tissues. 6.11 COLOR COMPENSATION MADE SIMPLE Spectral or color compensation has been treated extensively elsewhere [54-57]. Yet, it remains the most common source of error and vexation for cytometrists everywhere. This section is meant to be a practical guide to understanding and avoiding the pitfalls of color compensation, rather than a complete exploration of the physical principles and mathematics. Color compensation can be thought of as correcting for the spillover of a given fluorochrome from the fluorescence channel in which it is intended to be measured to all of the other fluorescence channels. The spillover coefficient from channel A into channel B (SCA-->B) (e.g., from the FITC channel into the PE channel) is measured by running a single-stained sample meant to be detected in channel A and dividing the geometric MFI of the peak detected in channel B (FIB) by that of channel A (MFIA) (Figure 6.11): _ MFIB (SCA-->B) - MFIA This spi1I0ver coefficient, SCA-->B' gives the proportion of the signal detected in channel A that will be spuriously detected in channel B. It is dependent not only on the emission spectrum of the particular fluorochrome, but also on the relative gains Understanding Clinical Flow Cytomelry 203 DJ[][][[JL FITC PE ECO PE-Cy5 PE-Cy7 FIGURE 6.11 Determining spillover coefficients. FITC-Iabeled Calibrite beads were acquired and the geometric MFI was determined for each ftuorescence parameter. The x-axes are labeled corresponding to the ftuorochromes to be measured in the experimental setup. The spillover of FTTC into the PE channel can be expressed as the MFI of FITC beads in the PE channel divided by the MFI of FITC beads in the FITC channel. In this example, it is 12.2%. The greater the difference in emission wavelength between FITC and the other ftuorochromes, the lower the spillover. This is not the case for tandem dyes, which often have significant spillover at the emission wavelength of the donor dye. of the two PMTs. Spillover coefficients can also be conceived of and calculated as slopes, since SCA....Bis the change in MFIBas a function of MFIA • In a two-color instrument, where spillover and compensation coefficients are almost identical, the fluorescence intensity of an event detected in channel B can be corrected for spectral overlap (FIBComp) by subtracting the fluorescence intensity detected in channel A multiplied by the spillover coefficient of A into B: The two-way spillover coefficients (A into Band B into A) are minimized, and thus compensation is minimized, when both PMTs are balanced. PMT balance cannot be achieved by applying equal voltages to the two PMTs, because PMT sensitivity is not standardized and often very different voltages are required to achieve equal results from different PMTs. Further, perfectly balanced PMTs are not always desirable. If you wish to detect a dim signal in channel B, the gain of PMTBcan be increased relative to that of A. This will result in greater spillover of A into B (and hence greater compensation) and less spillover of B into A (less compensation). This may seem counterintuitive, but consider how we defined the spillover coefficient. Similarly, it may be necessary to lower the sensitivity of a PMT if it is used to detect an exceptionally bright fluorochrome. It is for these reasons that PMT target channels cannot be chosen solely by evaluation of unstained controls, and it is often necessary to have different PMT target channel sets for different applications. One more item remains to be addressed before we