Cytometry (Communications in Clinical Cytometry) 38:161 175 (1999) Four Color Compensation Carleton C. Stewart* and Sigrid J. Stewart Laboratory of Flow Cytometry, Roswell Park Cancer Institute, Buffalo, New York Four-color immunophenotyping can now be routinely performed using either a single laser or dual laser flow cytometer. When a single laser instrument is used, the fluorochromes evaluated are usually FITC, PE, PE-TR and PE-CY5 (or PerCP). For two-laser excitation APC is generally used in place of PE-TR. Since each tandem dye construct contains PE, three of the four detectors are affected and compensation can be problematic. In this report we show that each tandem conjugated antibody, whether different batches from the same supplier or conjugates from different suppliers all require unique compensation. This inconsistency results in erroneous data, negates the use of single labeled particles as a method for providing adequate compensation and requires dual and triple labeled cells of known pattern to verify compensation. It is also shown that improper compensation can reduce or eliminate completely the detection of fluorescence emission from PECY5 conjugated antibodies. These problems are caused by a variation in energy transfer between PE and either TR or CY5 because the chemistry involved in preparation and conjugation to antibodies is not sufficiently controlled to produce reagents with uniform compensation requirements. The variation in tandem dye compensation can be addressed by either using the same tandem conjugated antibody, by using the same second step tandem reagent to an appropriate first step antibody or by using software compensation. The latter provides an easy solution because a unique compensation matrix can be produced for each antibody tandem conjugate. Cytometry (Comm. Clin. Cytometry) 38:161 175, 1999. 1999 Wiley-Liss, Inc. Key terms: flow cytometry; instrument and software compensation; tandem dyes; antibodies; single laser; dual laser Multiparameter flow cytometry is a powerful tool for the objective measurement of antigen co-expression on populations of hematopoietic cells (1 6). Over the last decade, we have learned that, by combining two antibodies together, patterns emerge in the form of cell clusters that represent the expression of antigens on the cell s surface. When three antibodies are added together, even more clusters can be resolved representing additional subsets of cells (7). Further increasing the dimensionality of the data by adding four antibodies not only results in increasing the resolving power for important subsets, but can also markedly decrease the cost of specimen processing by reducing the number of tubes analyzed and reducing the number of redundant antibodies used. Each time another reagent is added, the complexity of the analysis process is increased. For example, the potential for antibody co-blocking (8,9), or fluorochrome quenching (10), or energy transfer (11) by epitopes in close proximity must be thoroughly explored before accepting any data as valid. It is now possible to routinely perform four-color immunophenotyping using either a single (Beckman Coulter XL) or dual (BDIS FACSCalibur) laser instrument. When a single laser instrument is used, the most frequently employed fluorochromes, all excited at 488 nm by an argon laser, are FITC, PE, PE-Texas-Red (PE-TR) and PE-CY5. PerCP conjugated antibodies can be used instead of PE-CY5 conjugated ones. Currently, a limited number of antibodies directly conjugated with PE-TR in the form of ECD are available from Beckman Coulter. Otherwise, biotinylated or unconjugated antibodies can be used in combination with the second reagents conjugated with PE-TR, products available from Life Technologies. When a dual laser instrument is used, three fluorochromes (usually FITC, PE and PE-CY5 or PerCP) are excited at 488 nm by an argon laser and one fluorochrome (usually APC) is excited at 635 nm by a diode or HeNe laser. A large selection of antibodies is available directly conjugated with APC from many reagent suppliers. Grant sponsor: New York State Department of Health; Grant sponsor: National Cancer Institute (NCI); Grant numbers: #2 P30-CA16056 and #5RO1CA60200006. *Correspondence to: Carleton C. Stewart, Ph.D., Laboratory of Flow Cytometry, Roswell Park Cancer Institute, Elm and Carlton Streets, Buffalo, NY 14263. E-mail: Stewart@sc3101.med.buffalo.edu Received 8 January 1999; Accepted 6 April 1999 1999 Wiley-Liss, Inc.
162 STEWART AND STEWART Compensation represents a special problem for fourcolor flow cytometry because PE is present in two or three of the four fluorochromes used. While it is customary to provide compensation standards using cells separately labeled with each fluorochrome, this approach is not adequate for four-color compensation adjustments. When each fluorescently labeled cell is interrogated, compensation signals will be present at the appropriate electronic summation junctions for the multiple compensations required. Because only single labeled cells are used, only one color at a time is present at these junctions, and the actual effect of one fluorochrome on another will not be seen. Even if a mixture of the four populations in a single tube is analyzed, each color is interrogated one at a time because each single labeled cell appears in the laser beam at a different time, and the effect seen will still be as if they were run separately. Only when the same cell is labeled with two or more fluorochromes can the true compensation requirement be evaluated and the effect of compensating one fluorochrome on another fluorochrome appreciated. This means that control cells must be labeled with antibodies whose binding to the cells within the suspension is known. In this report, we show that the fluorochrome emission from PE-TR in the single laser configuration or PE-CY5 or PerCP fluorochrome emission in the dual laser configuration is exquisitely sensitive to improper instrument compensation. Because both of the tandem constructs (but not PerCP) can exhibit significant variations in their amount of PE leakage, proper compensation is unique for each batch of fluorochrome and antibody to which it is conjugated. This problem can produce significant logistic problems in acquiring correctly compensated data when samples contain different antibodies conjugated with the tandem dyes. A method for verifying the correct instrument compensation is provided. In addition, we provide a strategy, using software compensation that can be used to easily solve the compensation problems associated with the variation in tandem fluorochromes conjugated to different antibodies. MATERIALS AND METHODS For these studies peripheral blood from healthy donors was used to provide daily assessment for immunophenotyping and instrument performance (12,13). The use of excess human products for research purposes is performed using approved protocols by the Institutional Review Board at RPCI. Blood was collected in heparinized vacutainers (BD) and the cellularity was determined using a Coulter M430 hematology counter. One part of blood was washed with 14 parts of PBS containing 10 units of heparin per ml. Throughout this report a wash is defined as centrifuging the cells for 3 min. at 1500 G, aspirating the supernatant and resuspending the pellet in residual buffer. The cells were washed a second time with 14 volumes of PBS without heparin. This procedure is used to reduce platelet aggregation by removing clotting factors and eliminating immunoglobulins or soluble cell surface antigens to which selected antibodies may be directed. The cells were resuspended in PBS for the experiments, at a concentration of 5 10 10 6 cells per ml. The antibodies used for these studies were conjugated with fluorescein isothiocyanate (FITC), phycoerythrin (PE), PE-Texas Red (ECD), PE-cyanin 5 (PE-CY5), peridinin chlorophyll protein (PerCP) or allophycocyanin (APC). FITC-CD3, FITC-CD45, PE-CD3, PE-CD4, PerCP-CD45 and APC-CD45 were obtained from Becton Dickinson Immunocytometry Systems (BDIS; San Jose, CA). PE-CD19, PE- CD64, PE-CY5-CD2, PE-CY5-CD3, PE-CY5-CD8, PE-CY5- CD10, PE-CY5-CD22, PE-CY5-CD33 and PE-CY5-CD64 were obtained from CALTAG Laboratories (Burlingame, CA). PE-CY5-CD19 was obtained from Immunotech, Beckman- Coulter (Hialeah, FL). All antibodies were titered before use as previously described (14). Cells were stained as previously described in detail (15,16). Briefly, antibodies in the appropriate combinations in a final volume of 30µl were added to 10 75 mm polypropylene tubes (Falcon #2054) followed by 50µl of cells. The final volume of the sample (antibodies plus cells) was rigorously maintained at 80µl. Cells were incubated 15 min. in an ice bath and then 3.5 ml freshly prepared lysing reagent (4.13g Ammonium Chloride (NH 4 CL) -- Sigma A-5666, 0.5g Potassium Bicarbonate (KHCO 3 ) -- Sigma P-4913, 0.0185g Tetra Sodium EDTA - Sigma ED4SS) pre-warmed to 25 C was added and incubated 5 min. After centrifuging at 1500 g for three minutes, cells were washed with 3.5 ml PBS and fixed in 300µl of 2% formaldehyde (1:5 dilution in PBS of Poly- Sciences #04018 Ultrapure EM Grade 10% formaldehyde) for a minimum of 30 min., but no longer than 72 hours, prior to data acquisition. A FACScan (BDIS) was modified to acquire four-color data using a single laser (17). Briefly, a fourth PMT with associated electronics was added. Instrument compensation was made possible for all combinations of detectors, including those not available on a commercial FACScan. Instrument compensation could also be turned off so the user has a choice between hardware and software compensation. Data acquisition for the fourth color was made possible by using the DDE pulse processing capability of Lysys II Software and assigning the area parameter to FL4. A FACSCalibur dual laser flow cytometer (BDIS) was used for all the two laser excitation studies. Both instruments underwent a quality assessment each morning using DNA check microspheres (BC part #6603488). The acceptable variation in the peak channel was less than 5% and the coefficient of variation (CV) was less than 3% for all parameters except side scatter where the CV was less than 5%. Interlaser time for the FACSCalibur was calibrated at least three times per day according to the manufacturer s procedure, and the sheath fluid level was never allowed to decrease past one-third full. We have noticed instability in interlaser timing, which could be completely resolved, if the sheath tank is never allowed to decrease beyond one third full. When the instrument is functioning properly the timing verification passes on the first try, using the software supplied. If repeated tries are necessary, there is a timing problem that must be found
FIG. 1. Lymphogated cells for compensation. A lymphogate was applied to data from processed human blood to provide cells for compensation standards. and fixed. Usual causes for timing instability are low sheath tank volume, a partially blocked flow cell, high debris count, the red laser requires replacement or air bubbles in the sample stream. Preliminary instrument compensation was performed using cells stained with either FITC-CD45, PE-CD4, ECD-CD4 or APC-CD45 and with each reagent conjugated with PE-CY5 or PerCP. For each sample, 10,000 ungated events were acquired using a FSC threshold set at the extreme left side of the lymphocyte cluster (see Fig. 1). For the forward scatter gain settings we use, this threshold was 200. Acquired data was transferred to a PC based computer using Apple Talk FTP, where it is permanently archived on CD-ROM discs. The data was analyzed using WinList version 4.0 (Verity Software House, Topsham, ME). This program provides software compensation using matrix algebra and has been previously described (18). A lymphogate was used for analysis of all reagents except CD33 and CD64 where monocytes were also included in the gate and CD10 where granulocytes were also included. RESULTS For these studies, a lymphogate was used, as shown in Figure 1. Lymphocytes are ideal because they represent a relatively homogeneous population of small cells that exhibit low autofluorescence. Using single laser excitation and cells stained with only one antibody prior to mixing them together, the instrument compensation is shown in Figure 2 for the six possible bivariate displays of FITC versus PE, ECD versus PE, PE-CY5 versus PE, FITC versus ECD, PE-CY5 versus ECD and PE-CY5 versus FITC. Cells stained with each reagent are wholly contained within their appropriate single color region. One would expect that this instrument is properly compensated. Using the combination PE-CD8, ECD-CD4 and PE-CY5-CD2, to stain lymphocytes, all CD4 and CD8 positive cells should be CD2 and CD4 and CD8 positive cells should be mutually exclusive populations. As shown in Figure 3, the instrument is not properly compensated (top row); because some CD8 or CD4 cells are CD2 - and some CD8 cells are CD4 dim. Both of these problems can FOUR COLOR COMPENSATION 163 be corrected, as shown in Figure 3 (bottom row) after making an adjustment to the compensation, shown in the Figure 3 caption. It is not possible to see this change when a single labeled compensation standard is used because the acceptable limits are too broad (11.5% vs 9%). Beckman/ Coulter provides reagent combinations called Cyto- Comp reagent kit (part # 6607021) for use with their XL flow cytometer for verifying compensation. They also produce PE-TR (ECD) tandem conjugated antibodies, with consistent compensation requirements, but the number of reagents is limited. It may still be necessary to stain cells with the current batch, of each reagent combination, to verify that the Cyto-comp kit behaves the same way as the reagents being used to stain cells. This verification would only be necessary when new batches of reagents are initiated. Instead of using instrument compensation, software compensation, such as that provided by Beckman Coulter with their XL flow cytometer or WinList by Verity Software House, can be used. When using WinList software compensation, single color compensation standards are prepared as described above and uncompensated files collected. This program uses matrix algebra previously described in detail (18) for determining compensation. Nevertheless, a multicolor compensation standard like that shown in Figure 3 must still be evaluated to obtain the correct compensation in software. When using software compensation, the samples are stained with the desired antibody combinations and the data is acquired uncompensated, as shown in Figure 4, top row and the compensation matrix applied to this data to produce the compensated data shown in the second row. For comparison, the bottom row shows the same specimen re-run after instrument compensation. These results clearly show that there is no difference in the patterns produced between instrument and software compensation. The percentages of positive cells for software versus instrument compensation, compared in Table 1, are in good agreement. One feature of software compensation is that the matrix can be named so that it can be incorporated into a data analysis macro. In this way a unique compensation matrix which can be automatically associated with its unique fluorochrome antibody conjugation batch. This capability provides for custom compensation described in more detail later when using the four-color, two laser applications. The most serious problem associated with either three or four-color immunophenotyping is the variation in compensation required for the tandem complexes. This variation is illustrated in Figure 5, and it is due to the inability to produce a consistent tandem of either PE-TR (not shown) or PE-CY5 that has exactly the same amount of PE leakage. When a second laser is used to excite APC, the CY5 in the PE-CY5 tandem is also excited. Since the CY5 compensation will differ from the APC compensation, this additional problem must also be appropriately dealt with. PerCP reagents are also excited to a lesser extent by the second laser. As shown in Figure 5 and Table 2, the PE-CY5-tandem variations cover a range from 0 to 8.9% for FL2-%FL3 and
164 STEWART AND STEWART FIG. 2. Instrument compensation. Blood was aliquoted into four tubes which contained either FITC-CD45 (FL1), PE-CD4 (FL2), PE-CY5-CD8 (FL3) or ECD-CD4 (FL4). After processing, data was acquired on a FACScan modified to collect four colors of fluorescence and the compensation adjusted so each population was within the appropriate quadrant. The six-parameter combinations of bivariate data are displayed. Compensation settings (in percent) are FL1 - %FL2 0.5, FL2 - %FL3 12, FL2 - %FL4 26, FL3 - %FL4 15, FL4 - %FL1 0.5 and FL4 - %FL3 50. The PMT high voltages were FL1 675, FL2 600, FL3 715 and FL4 700. from 16% to 42% for FL4-%FL3 on our FACSCalibur. These percentages will vary depending on the high voltage adjustments for each PMT and the batch of PE-CY5 used. When two different batches of CD8 were compared, batch 801 exhibited 1.1% compensation for FL2-%FL3 and 35.8% for FL4-%FL3. The mean values for this batch are shown in Table 2. Batch 901 (not shown in Table 2) exhibited a similar FL2-%FL3 compensation, but a much higher FL4- %FL3 compensation of 61.3%. A second batch of PE-CY5- CD19 exhibited a similar variation to the batch shown in Table 2. For this batch FL2-%FL3 was 1% and FL4-%FL3 was 39.1%. Antibodies conjugated with PerCP, Table 2, required no compensation for FL3-%FL2 and 3% compensation for FL4-%FL3. No variation in compensation requirements was found among different antibodies conjugated with PerCP. The problems illustrated in Figure 6, associated with using cells (or microspheres) stained with a single color for obtaining the correct compensation that were described for a four-color single laser instrument are also found with a two-laser instrument. As shown in the top row, four PE-CY5 reagents were compensated using single antibody stained cells. The amount of compensation for each reagent is reported in the caption. For each reagent the x axis fluorescence intensity (also shown in the view) is much higher than that found for the same reagent in row 2 when the PE-CY5 conjugated antibody was combined with APC-CD45. Indeed, very few CD8 T-cells and no CD64 monocytes were found. When this same combination is prepared with FITC-CD45, as shown in the third row, the PE-CY5 conjugated antibody fluorescence intensity is identical to that found when it is evaluated as a single reagent (top row). As shown in the fourth row, this situation does not occur if each antibody is conjugated with PE, rather than a PE-CY5 and then combined with APC-CD45. This problem illustrated in row 2 is unique to the combination of PE-CY5 conjugated antibodies combined with APC-CD45. It is caused by overcompensation of FL3-% FL4 when using a single color compensation sample. So why is a single color compensation sample inadequate? When a single color is used for compensation, there are no other signals at the summation junctions within the compensation circuits except those from one fluorochrome. Even if cells stained separately with different fluorochromes are mixed together, they appear one at a time for interrogation. When cells are dually positive, fluorescence signals from each fluorochrome will appear at the summation junctions simultaneously. Since the first fluorochrome will have a component of the second s color and vice versa, each of them will produce a different compensation requirement when evaluated on doubly stained cells compared to singly stained cells. The degree of difference in the compensation settings between the single versus double labeled cells will depend on the intensity of each fluorochrome and the percent of their respective
FOUR COLOR COMPENSATION 165 FIG. 3. PE-TX-Red & PE-CY5 compensation. Blood was processed using the antibody combination PE-CD8, ECD-CD4 and PE-CY5-CD2 (TC). The top row shows the pattern obtained after single color compensation. The appropriate adjustments were made to the compensation to obtain the correct pattern shown in the bottom row. All compensation and high voltage settings are listed in the caption for Figure 2 except FL3 - %FL2 and FL3 - %FL4. These are the most sensitive compensation parameters. The compensation for FL3 - %FL2 for TC-CD2 vs PE-CD8 in the column 1, top view using single labeled cells was 11.5% and for column 1, bottom view using double-labeled cells was 9.0%. The compensation for FL3 - %FL4 for TC-CD2 vs ECD-CD4 in column 2, top view, using single labeled cells was 58% and for column 2, bottom view, using double-labeled cells was 40%. spectra that overlap. (A detailed description of why there is a difference may be found in the appendix.) This means that a dual stained cell preparation must be used to verify compensation, as shown in Figure 7. To obtain the correct compensation, a very bright APC antibody, such as APC-CD45 that will stain all cells, is combined with a PE-CY5 or PerCP antibody, such as CD8, that stains only a subset. As illustrated in Figure 7-A, the correct compensation is obtained by adjusting for the maximum separation of the negative and the brightest stained clusters (shown by the line). This is the same position for single labeled CD8 positive cells, shown in Figure 6. CD8 is a good antibody to use because both bright and dim populations of CD8 cells are present and each should be well resolved. In Figure 7B, the sample has been overcompensated by only 4%, and the CD8 appears to be heterogeneously expressed, the cells are not as bright and the dim cells are poorly resolved as they merge with the bright cells. In Figure 7C, the sample is not compensated and the bright (b), dim (d) and negative (n) populations are still resolved. When the dual stained standard is used, the optimal separation between positive and negative cells can be obtained. Once adjusted, the setting is good for any PE-CY5/APC antibody combination. Unfortunately, even the two-color combination is inadequate and a three-color compensation standard is needed because PE-CY5 emission occurs at the summation junctions of FL2, FL3 and FL4. By combining mutually exclusive PE (CD4) and PE-CY5 (CD8) bright antibodies with a bright APC antibody coexpressed (CD45) by both, the optimal separations for all populations can be achieved through compensation. The reasons why this is necessary are illustrated in Figure 8. When a single color sample such as PE-CD4 is used for compensation (top row of Fig. 8) PE (FL2) vs PE-CY5 (FL3) (A) and PE (FL2) vs APC (FL4) (B) and APC (FL4) vs PE-CY5 (FL3) (C) look like they are correctly compensated. When cells stained simultaneously with three bright antibodies such as PE-CD4, PE-CY5-CD8 and APC-CD45, are evaluated, as shown in the second row, PE-CD4 (FL2) (D) is clearly under compensated and there are a high number of apparently CD8 dim cells (F) shown in the ovals. This
166 STEWART AND STEWART FIG. 4. Comparison of compensation methods. Blood was processed with the four-color antibody combination FITC-CD3, PE-CD8, ECD-CD4 and PE-CY5-CD2 (TC). Data was acquired using no compensation (top row), software compensation (middle row) or instrument compensation (bottom row). The six-parameter combinations of bivariate data are displayed. See Figure 2 and Figure 3 captions for high voltage and compensation settings. Table 1 Comparison of Cluster Event Count in Software Compensated and Instrument Compensated Data Cluster a % of gated cells compensated CD3 CD8 CD2 CD4 Software Instrument 13.4 12.6 8.7 8.2 4.7 3.9 15.9 19.5 51.6 50.9 2.5 2.4 a Populations 1% not included. appears to be caused by under compensation of either FL3-%FL2 or FL3-%FL4. The tendency might be to increase compensation for FL3-%FL2 to correct the under-compensation seen in Figure 8D. When this is done, and the single stained cells are reevaluated, third row, PE-CD4 stained cells (G) are correctly compensated, but they are now under-compensated for FL4-%FL2 (H, I). Clearly the selection of FL3 - %FL2 was the wrong choice. If, instead, FL3-%FL4 is adjusted and the single stained cells reevaluated (J, K, L) as well as the multicolor stained samples, M, N, O correct compensation has been achieved and verified. When performing this compensation make sure that the PE-CY5 positive cells are in the same location as they were in Figure 7 (vertical line) i.e. the maximum separation has been achieved. Therefore, it is necessary to evaluate a three-color stained sample to verify correct compensation. This antibody combination can be evaluated every time four-color samples are evaluated to verify that the instrument is properly compensated. We have defined two different problems. The first problem is encountered when each TC reagent requires a different amount of compensation and the second is encountered when we try to accurately adjust compensation using single color standards. What is a simple solution to these problems? Because each PE-CY5 conjugated antibody and any PerCP conjugated antibody requires its own unique compensation and it is desired to use several different PE-CY5 antibodies and PerCP antibodies in separate combinations with other antibodies that will be evaluated on the same specimen, we recommend using software compensation. This provides for a compensation matrix unique for each PE-CY5 antibody that has been combined with three other reagents. When a unique compensation matrix is generated in software for each
FOUR COLOR COMPENSATION 167 FIG. 5. Variation in compensation for PE-CY5 reagents. Separate samples are processed with PE-CY5 conjugated CD2, CD19 or two different batches of CD8. The data were acquired using no compensation or after adjusting the compensation to the value shown. The high voltage was FL3 690V, FL4 730V. Table 2 Variation Among and Stability of Compensation Requirements for Third Color Reagents FL2-%FL3 FL4-%FL3 a Antibody Supplier Mean % STD Mean % STD PE-CY5-CD2 Caltag 8.9 4.4 19 2.2 PE-CY5-CD8 Caltag 1.0 15.0 41 2.9 PE-CY5-CD10 Caltag 0.9 12.6 26 1.1 PE-CY5-CD19 Immunotech 1.1 9.4 18 1.1 PE-CY5-CD22 Caltag 0.0 0.0 35 2.6 PE-CY5-CD33 Caltag 0.0 0.0 42 2.4 PerCP-CD45 BDIS 0.0 0.0 3 5.5 PE-CY5-CD62L Caltag 1.5 9.5 34 3.0 PE-CY5-CD64 Caltag 0.0 0.0 16 2.7 a The amount of compensation is dependent on the high voltage and may vary markedly from those shown here and still be correct. For this experiment, the high voltages were FL2 640V, FL3 690V and FL4 730V. The mean and % STD for each reagent measured monthly for seven months is shown. combination, it can be applied to uncompensated acquired data at the time the list mode files are analyzed. Is it necessary to create a new compensation matrix every time (e.g. daily) that a PE-CY5 conjugated antibody is evaluated? We determined the stability of 8 separate PE-CY5 antibodies and a PerCP antibody over a period of seven months by staining cells monthly and recorded the compensation required. The results, summarized in Table 2, demonstrate that compensation for each antibody is stable and varies 5% for FL4 - %FL3 when properly stored. Variation was greater than 5% for FL2 - %FL3, but this is because so little compensation is needed, that a change in only 0.1% would make a large statistical difference. Light exposure, however, can break down the PE-CY5 tandem, causing increased PE leakage and this would invalidate the matrix. Such an occurrence, however, would show up during periodic evaluation, as controls would fall outside acceptable thresholds established for each PE-CY5 antibody. This procedure offers a means for quality assessment of stored reagents on a routine basis. To perform the quality assessment, for both maximum separation of negative and bright clusters and the unique compensation for (FL3-%FL4), we suggest staining two groups of cells. The first group is stained with each PE-CY5 antibody alone. The second group is one tube of cells that have been stained with the antibody combination PE-CD4, PE-CY5-CD8 and APC-CD45. All three markers are very bright. A different combination might be chosen, but the important criteria is that all are bright markers, FL2 and FL3 cells are all FL4, but FL2 cells do not co-express FL3. Each PE-CY5 antibody stained cells is mixed with the cells stained with the combination. The pattern generated is shown in Figure 9. In the top row, CD4, CD8 and CD45 are properly compensated and maximum separation of CD45 CD4 and CD45 CD8 clusters from those cells that are only CD45 is achieved. TC-CD2 cells have not yet been properly compensated. Both FL2-%FL3 and FL4-
168 STEWART AND STEWART FIG. 6. Over compensation of PE-CY5 and APC-CD45. Blood was stained with each antibody combination shown on the axis. Each of them was compensated using the single stained cells shown in the top row. FL4 - % FL3 for PE-CY5-CD19 was 32.8%, PE-CY5-CD8 was 50%, PE-CY5-CD3 was 31.5% or PE-CY5-CD64 was 28.6% (for this antibody, monocytes were included in the gate) as a single antibody (top row) or combined with APC-CD45 (second row), or FITC-CD45, third row. In the fourth row, PE-CD19, PE-CD8, PE-CD3 or PE-CD64 was combined with APC-CD45. Note that PE-CD8 is brighter than PE-CY5-CD8, PE-CD3 is brighter than PE-CY5-CD3 and PE-CD64 is very much brighter than PE-CY5-CD64. The high voltage settings were FL1 690V, FL2 640V, FL3 690V and FL4 730V. FIG. 7. Using a double labeled compensation standard. Blood was stained with PE-CY5-CD8 and APC-CD45. When properly compensated (A, FL3 - %FL4 20%), the maximum spatial separation of each cluster representing CD8 negative (n), dim (d) and bright (b) cells, occurs. Over compensation (B, FL3 - %FL4 24%) significantly reduces the spatial separation and produces erroneous data (note position of cells relative to the vertical line in A vs B). In C, no compensation has been applied, but the three populations can still be seen. Proper compensation can only be achieved using cells stained with both antibodies so the maximal separation of n and b can be determined. The high voltage settings were FL3 690V and FL4 730V. %FL3 are adjusted to properly compensate the CD2 cells and the pattern shown in the bottom row is found. The instrument has been properly compensated for PE-CD4, APC-CD45 and TC-CD2. In this paradigm, instrument compensation is set for all antibodies except the TC reagents. For these the unique compensation matrix for the TC reagent is determined and subsequently used for software compensation.
FOUR COLOR COMPENSATION 169 FIG. 8. Using a Triple Labeled Compensation Standard. Blood was stained as a single label with PE-CD4, PE-CY5-CD8 or APC-CD45. Unstained cells were used to set the high voltage which was FL2 640V, FL3 690V and FL4 730V. The compensation was adjusted to FL2-%FL3 1%, FL3-%FL2 12.4%, FL3-%FL4 18% and FL4-%FL3 41%. A, B and C show that PE-CD4 cells appear to be properly compensated. Cells were also stained with the combination containing PE-CD4, PE-CY5-CD8 and APC-CD45. D, E and F show the bivariate histograms of these cells. FL3-%FL2 was increased to 16.3% using the mutually exclusive combination of PE-CD4 vs PE-CY5-CD8 (D). G, H and I show the reacquisition of PE-CD4 at this compensation setting. FL3-%FL2 was set to the original setting (12.4%) and FL3-%FL4 adjusted to 20% (from 18%) and J, K and L show reacquisition of PE-CD4 at this compensation setting and M, N and O show reacquisition of the triple labeled compensation standard.
170 STEWART AND STEWART FIG. 9. Verification of compensation. Blood is processed in two tubes. The first tube contains the antibody combination PE-CD4, PE-CY5-CD8 and APC-CD45 and the second tube contains only PE-CY5-CD2. After fixation, the contents of both tubes were combined and data acquired. Since the PE-CY5-CD8 requires little compensation for FL2-%FL3, it is used in the three-color antibody combination for insuring maximum separation of CD45 versus FL3. PE-CY5 reagents stained separately and mixed will appear to be uncompensated for FL4-%FL3 (A, B, C). When properly compensated, either in software or hardware, the pattern shown in D, E, F will appear thereby verifying correct compensation has been achieved. In this example FL2 - %FL3 10 and FL4 - %FL3 20 at high voltages for FL2 640V, FL3 690V and FL4 730V. Each PE-CY5 is verified monthly and whenever a new batch of reagent is introduced. DISCUSSION These results show that every PE-CY5 tandem requires its own unique compensation, but the stability for any single batch is quite good. PerCP, a natural tandem reagent does not exhibit this variation, as all PerCP conjugated antibodies require the same low compensation, nor does the 635 nm laser efficiently excite it. Unfortunately it is not as bright as the PE-CY5 reagents. The variation in compensation of PE-CY5 reagents is caused by both the leakage of PE photons due to a variation in energy transfer to CY5 among different tandem batches as well as conjugation to each antibody and to the excitation of CY5 in the tandem by the diode laser. Some suppliers of TC reagents have controlled their product s compensation for PE-%TC compensation to about 1%; others supply TC reagents that may require 5 10% PE-%TC compensation. For any single TC reagent, its compensation requirement is stable providing the reagent is not exposed to light during storage. The amount of TC found in the APC channel is much higher and this is because the CY5 is excited by the 635 nm laser. While the exact physical chemical relationships may not have been elucidated, the basis for this effect is the inability to produce conjugations of PE-CY5 to antibodies that are identical each time. The amount of PE-%TC and APC-%TC compensation for a given construct will depend on the tandem chemistry, the conjugation chemistry of the tandem to the antibody and the conjugated antibody s binding to its epitope on the cell. All three categories are variable with each preparation and that causes an effect on the efficiency of energy transfer from PE to CY5, which affects compensation. The variation among reagents creates a practical logistic problem in evaluating specimens when multiple tandem conjugated antibodies are combined with other antibodies in a multiple tube screen. The problem can be solved in three ways: 1. Always use the same PE-CY5 reagent. Since only one tandem is used, there is no variation, and the same compensation can be used for all samples regardless of the other three antibodies combined with it. For example, CD45-PE-CY5 (or PerCP) might be chosen as the common reagent. The problem with this solution is that the
investigator no longer has the flexibility to use other tandem antibodies in the four-color combinations. 2. Use a biotinylated antibody and a PE-CY5 avidin. An unconjugated antibody and a second antibody conjugated with the tandem could also be used. Since the tandem avidin or second antibody is always the same, there is no variation in compensation. Thus, the investigator now has the choice of any antibody in the four-color combination, but a second step is required. 3. Use software compensation. Using WinList, version 4 (or 5), a compensation matrix unique to each PE-CY5 or PerCP conjugated antibody can be produced. The appropriate compensation matrix is applied to each unique antibody among the combinations in the screen. This problem does not occur with PerCP reagents, as all of them will require the same FL4-%FL3 compensation and they require no FL2-%FL3 compensation. Nevertheless, both FL4-%FL3 and FL2-%FL3 compensation will be different for each PE-CY5 reagent and between the PE-CY5 and PerCP reagents. Over or under compensation is a problem that must be recognized by the laboratorian. A small change in the percent compensation can make a well-resolved cluster become heterogeneous in shape as shown in Figure 7. Cells with low epitope density that renders them dimly fluorescent may become completely negative, Figure 6, row 2. To insure that each parameter is properly compensated, cells should be evaluated every time data is collected using antibody combinations that will allow for verifying that the maximum difference between positive and negative cell clusters has been achieved. An example of how to accomplish this was shown in Figures 7, 8 and 9. For software compensation, the files are used to validate the compensation matrix. The compensation problems identified here will be even more acute in the future as laboratorians make their own combinations using analyte specific reagents (ASRs). It is likely that the manufacturers will introduce fewer FDA approved combinations (in vitro diagnostic device products), because it will be much less expensive for them to provide ASRs. All clinical laboratories performing flow cytometric evaluations using ASR combinations prepared by them must implement the regulations, which went into effect on November 23, 1998. Since all compensation combinations are available in software, it is important that the correct ones are used. It is also important to realize that a population of cells along one of the six pairs of axes may be a projection of an uncompensated population and not one that is actually positive for the antibody. All six bivariate plots of four-color data must be viewed before determining which pair requires compensation, because the bivariate that is not shown may be the one that actually shows the proper parameters to compensate. Only the relevant bivariates have been shown in this report to save space. The importance of choosing the proper parameters is best illustrated when PE, PE-CY5 and APC antibodies produce bright cellular staining such as the combination of PE-CD4, PE-CY5-CD8 and APC-CD45. As was shown in FOUR COLOR COMPENSATION 171 Figure 8, after compensation using cells stained with only a single antibody (top row), it looked as though the instrument was correctly compensated. But when cells stained with all three combined antibodies were evaluated, a high frequency of CD4 dim, CD8 cells were found. This situation will only occur when the FL4 cells are also FL2. If extra compensation is added to FL3 - %FL2 to correct this problem, a summation occurs that actually produces under compensation of CD4 vs CD45 (Figure 8H). Correct compensation can be achieved by increasing FL3-%FL4 (Figure 8K). This illustrates the interaction of the compensation circuits with one another and the need to understand how they work, so that the correct parameters are compensated. A discussion of how compensation works may be found in the Appendix. We recommend software compensation because a matrix can be prepared for each antibody combination and then applied to the sample during analysis of the data. Uncompensated data can be retrospectively compensated in software, but over compensated data obtained using hardware compensation is lost forever. Using WinList,a combination specific macro can be written that will automatically bring up the correct matrix as the sample is analyzed. Providing the PE-CY5 reagents are properly stored and not exposed to bright light, the compensation matrix for each batch of any PE-CY5 reagent is stable for at least six months. Further, anytime a new batch of reagent is initiated, a new compensation matrix must be generated because it is highly unlikely compensation requirements will be the same. There are two software compensation approaches to consider. The first is to acquire all data uncompensated and do all the compensation in software. The advantage of this approach is that all data can be compensated retrospectively. A corollary to this approach would be to use instrument compensation for FL1 and FL2 because they are not affected by the problems encountered with the tandems. Alternatively, instrument compensation for all parameters as described in Figure 9, can be performed, except for FL2-%FL3 and FL4 - %FL3, which is performed in software for each PE-CY5 reagent and for PerCP reagents. While there is no advantage to this approach, it is perhaps more traditional because all the other compensation conditions are performed on the instrument. There are currently two software packages available that are capable of compensation, WinList and FloJo (19). We have no experience with FloJo. An interaction between the PMT voltage (gain) and the percentage of that signal appearing at the summation junction from a second PMT produce compensation. The percentage compensation is usually adjusted by changing the attenuation of the compensation network. There is a perception that a high percentage of compensation is too much and this is certainly true if it is 99% wherein the network is out of range. In practice, any desired percentage compensation can be set and the events brought into compliance by adjusting the appropriate PMT voltage. This procedure will insure that the compensation network (or software compensation matrix) is in
172 STEWART AND STEWART range. Thus, the FL4-%FL3 compensation requirement for CD8 (batch 901) could be reduced to 30% by increasing the FL3 PMT voltage or decreasing the FL4 voltage. (See Table 2) Adjusting the PMT voltage, however, will produce a change in the position of unstained cells. Four-color immunophenotyping significantly improves our ability to resolve unique subsets of cells within a complex population. To insure that the data has been properly acquired, the laboratorian must have an understanding of the entire process, including the instrumentation, so potential artifacts are recognized. While some may feel increasing the number of antibodies is too complex to pursue, this application continues to expand our understanding of subsets of hematopoietic cells and in reducing the cost of processing. Using appropriate combinations of antibodies it is possible to both identify cells and determine what they are doing. Never before has the acquisition of so much relevant data been possible using wellestablished procedures. The addition of even more antibodies in the combination (19) will further expand this knowledge base as we strive to understand the biology of hematopoiesis in health and disease. ACKNOWLEDGMENTS The authors wish to thank Mary Beth Dell and David Sheedy for their expert assistance in specimen processing and data analysis. LITERATURE CITED 1. Jennings CD, Foon KA. Recent advances in flow cytometry: application to the diagnosis of hematologic malignancy. Blood 1997; 90:2863 2892. 2. Roth G, Schmitz G. Consensus protocol for the flow cytometric immunophenotyping of hematopoietic malignancies. Working group on flow cytometry and image analysis. Leukemia 1996; 10:877 895. 3. Braylan RC, Borowitz MJ, Davis BH, Stelzer GT, Stewart CC. U.S. -Canadian consensus recommendations on the immunophenotypic analysis of hematologic neoplasia by flow cytometry: data reporting. Cytometry 1997; 30:245 248. 4. Davis BH, et al. U.S. -Canadian consensus recommendations on the immunophenotypic analysis of hematologic neoplasia by flow cytometry: medical indications. Cytometry 1997; 30:249 263. 5. Borowitz MJ, et al. U.S. -Canadian consensus recommendations on the immunophenotypic analysis of hematologic neoplasia by flow cytometry: data analysis and interpretation. Cytometry 1997; 30:236 244. 6. Loken M, Terstappen L, Civin C, Falker M. Flow cytometric characterization of erythroid, lymphoid and monomyeloid lineages in normal human bone marrow. In: O. Laerum, Bjerhnes R, editors. Flow cytometry in hematology, New York; Academic Press, 1992. 7. Stewart CC, Stewart SJ. Multiparameter analysis of leukocytes by flow cytometry. In: Darzynkiewicz Z, Robinson J, Crissman H, editors. Methods in cell biology, New York; Academic Press, Inc., 1994; 41: 61 79. 8. Ierino FL, Hulett MD, McKenzie IF, Hogarth PM. Mapping epitopes of human Fc gamma RII (CDw32) with monoclonal antibodies and recombinant receptors. J Immunol 1993; 150: 1794 1803. 9. Stein R, Belisle E, Hansen HJ, Goldenberg DM. Epitope specificity of the anti-(b-cell lymphoma) monoclonal antibody, LL2. Cancer Immunol Immunother 1993; 37:293 298. 10. Deka C, Lehnert BE, Lehnert NM, Jones GM, Sklar LA, Steinkamp JA. Analysis of fluorescence lifetime and quenching of FITC-conjugated antibodies on cells by phase-sensitive flow cytometry. Cytometry 1996; 25. 271 179. 11. Waggoner AS. Fluorescent probes for cytometry, In: Melamed MR, Lindmo T, Mendelsohn ML, editors. Flow cytometry and sorting, New York; Wiley-Liss, 1990; p 209 226. 12. Owens M, Loken M. Flow cytometry principles for clinical laboratory practice. Quality assurance for quantitative immunophenotyping. New York; Wiley-Liss, 1995. 13. Stewart CC, Stewart SJ. Phenotypic analysis. In: Robinson JP, Darzynkiewicz Z, Dean P, Dressler L, Rabinovitch P, Stewart C, Tanke H, Wheeless L, editors. Current protocols in cytometry, New York, John Wiley & Sons, Inc., 1997; 6.0.1 6.2.18. 14. Stewart CC, Stewart SJ. Titering antibodies. In: Robinson JP, Darzynkiewicz Z, Dean P, Dressler L, Rabinovitch P, Stewart C, Tanke H, Wheeless L, editors, Current protocols in cytometry, New York, John Wiley & Sons, Inc., 1997; 4.1.1 4.1.13. 15. Stewart CC, Stewart SJ. Cell Preparation for the identification of leukocytes. In: Darzynkiewicz Z, Robinson J, Crissman H, editors. Methods in cell biology, New York; Academic Press, Inc., 1994; 41:39 60. 16. Stewart CC, Stewart SJ. The use of directly and indirectly labeled monoclonal antibodies in flow cytometry. In: Davis WC editor, Methods in molecular biology. Humana Press, Inc., 1995; 45:129 147. 17. Timm EA, Podniesinski E, Duckett L, Cardott J, Stewart CC. Amplification and detection of a y-chromosome DNA sequence by fluorescence in situ polymerase chain reaction and flow cytometry using cells in suspension. Cytometry 1995; 22: 250 255. 18. Bagwell CB, Adams EG. Fluorescence spectral overlap compensation for any number of flow cytometric parameters. In: Landay A, Ault KA, Bauer RD, Rabinovitch, editors. Clinical flow cytometry, Annals of the New York Academy of Sciences, New York, 1993: 677:94:167 184. 19. Roederer M, De Rosa S, Gerstein R, Anderson M, Bigos M, Stovel R, Nozaki T, Parks D, Herzenberg L, Herzenberg L. 8 color, 10-parameter flow cytometry to elucidate complex leukocyte heterogeneity. Cytometry 1997; 29:328 339. APPENDIX The purpose of this appendix is to provide a description of how compensation works. This may be helpful in understanding of the results presented in this report. There are two compensation combinations for two-color, six for three-color and 12 for four-color. As the number of parameters increases, the complexity and cost of electronic compensation increases, therefore, software compensation becomes a more attractive alternative. Since higher order compensation is an array of the basic twocolor compensation, this discussion will be limited to two-color compensation. The emission spectra for two fluorochromes are shown in Figure A-1. The spectrum is an intensity (probability) distribution of emitted photons (intensity) of differing energies (wavelength or color). The parallel lines represent the colors that are determined by band pass filters for detection by the photomultiplier tubes (PMT and preamplifier). A simplified schematic representation of a compensation circuit is shown in Figure 2. The actual circuit is more complex for several electronic design reasons. In Figure A-2, FLA is a PMT and pre-amplifier with a band pass filter for selection of photons whose energy is the color green. Similarly, FLB is a PMT and pre-amplifier with a band pass filter for selecting photons whose energy is yellow orange. We will call it red for discussion. The pre-amplifiers amplify the PMT output for processing by the compensation circuits. Each of the fluorochromes emit some light that will pass through both optical filters. The amount of this light depends on the emission spectrum of the dye and the characteristics for the optical filters used for each PMT. Compensation corrects for the light that enters the secondary detector. Since this light is directly proportional to the light observed in the primary detector, it can be removed by subtracting a proportion of the primary signal from the secondary detector. This can be performed for multiple detectors simultaneously.
FOUR COLOR COMPENSATION 173 FIG. A-1. Emission Spectra of Two Fluorochromes. Emission of photons from fluorochromes excited by light occurs over a broad range of energies. This emission is described by an intensity distribution of photons at different wavelengths. The spectra of two fluorochromes may overlap one another and the photons from either that have the same wavelength (or color) are indistinguishable. Optical filtration, shown by the parallel lines representing the filters bandwidth, can be used to restrict the color that is transmitted through an optical detection system. Nevertheless, overlap of part of the unwanted fluorochromes occurs. When cells are labeled with a single color (Fig. A-2a), photons from the fluorescent cells are focused through optical filters onto both PMTs. FLA will produce a greater output signal for green photons than FLB used to detect the red photons because the band pass filters in front of each PMT selectively filters the appropriate color. The signal height at the output of each PMT/pre-amp will be proportional to the photon intensity of each selected color and this is measured in volts. There are two summation circuits, %A and %B in Figure A-2. The summation device is called a differential operational amplifier, OPAMP, which can process two signals by inverting one of them (at the negative input) and adding them together. For compensation, the signal from FLA goes to the positive input of OPAMP A. The output signal from OPAMP A is routed through an adjustable circuit %A, into the negative input of OPAMP B. This negative signal from OPAMP A is adjusted so that it is exactly equal to the height of the signal at the output of FLB, thereby canceling the unwanted signal in FLB. Thus, we have taken a percentage of signal A and subtracted it from signal B, often written, B - %A. Note: Compensation is a linear process. The signal may then be amplified through a logarithmic amplifier. In a similar manner, when the tube of cells is labeled only with the second color, whose primary detector is FLB (Fig. A-2b), FLB will produce a greater output signal than FLA. As before, the bandpass filters selectively determine the amount of colored light detected by each PMT. Now the FLB signal is routed to the positive input of OPAMP B. The output of FLB is routed through the adjustable negative circuit, %B, where the height of signal B is adjusted to the same height as the unwanted signal A. Thus, we have taken a percentage of signal B and subtracted it from signal A or A - %B, thereby canceling the unwanted signal FLA. An important, often overlooked, corollary to this description is the high voltage adjustment. This amount of subtraction depends upon the relative sensitivity of the two detectors to the two fluorochromes. This changes when optical filters are changed or when the high voltage is changed. If the high voltage on one of the PMTs is increased the signal heights in that detector will be proportionally higher. This will cause the signal height at the negative input of the OPAMP to be higher and unless it is readjusted, over-compensation will occur. Similarly, if the high voltage of the PMT is lowered, its signal height will be lower, causing the adjusted signal height at the negative input of the OPAMP to be lower and undercompensation will result. Thus, the compensation circuit could be adjusted to a fixed value and the cells actually compensated by adjusting the high voltage. The high voltage and compensation circuits work together; adjusting one will always affect the other. What is different when the same cells express both colors? There is a difference and it occurs for any color combination and becomes increasingly more problematic as the number of measured colors increases. The awareness of the problem has not been high because it is less noticeable when only two or three fluorochromes are used with fairly well separated emission spectra. To see quantitatively what is occurring, we can use voltage to represent signal height by letting green and red fluorescence each equal 5000 mv as illustrated in Figure A-3. Since each fluorochrome has a constant proportion of photons in each detector s energy range, 5% of the red fluorochrome might overlap the green fluorochrome and
174 STEWART AND STEWART FIG. A-2. Compensation for single labeled cells. A conceptual diagram of compensation for cells labeled with fluorescent dyes. FLA and FLB represent the combined PMT and pre-amp. The triangular devices are OPAMPs whose function is to invert any signal at the negative input and then add it to any signal at the positive input. In Figure A-2a, when only green cells are acquired, the signals from FLA will represent the green component of emission and from FLB the red component from the green fluorochrome. In Figure A-2b, when only red cells are acquired, the signals from FLA will represent the green component from the red fluorochrome and the red component is detected by FLB. FIG. A-3. Compensation for double-labeled cells. The component of each fluorochrome is represented by individual signal heights that are added together. The latter being what the detector actually measures. The higher first signal is from the fluorochrome primarily detected and the lower second signal from the overlapping component of the other fluorochrome. The signal processing is identical to that shown in Figure A-2.
FOUR COLOR COMPENSATION 175 Table A-1 OPAMP Input Voltages (ma) for Single-Labeled Cells OPAMP A (mvolts) OPAMP B (mvolts) Input Green cells 5000 A %B 50 1000 B %A 1000 Output 4950 0 Red cells 250 A %B 250 5000 B %A 50 Output 0 4950 Table A-2 OPAMP Input Voltages (ma) for Double-Labeled Cells OPAMP A (mvolts) OPAMP B (mvolts) Input Green 5000 A %B 250 1000 B %A 1000 Red 250 50 5000 50 Total 5250 A %B 262.5 6000 B %A 1200 Output 4987 4800 Table A-1 OPAMP Input Voltages (ma) for Single-Labeled Cells Input Output Green labeled cells OPAMP A 5000 A 5% B 0 5000 OPAMP B 1000 B 20% A 1000 0 Red labeled cells OPAMP A 250 A 5% B 250 0 OPAMP B 5000 B 20% A 0 5000 the compensation for it so that A -%B 5%. The green fluorochrome might overlap the red fluorochrome by 20% sothatb-%a 20%. (The actual percentage of compensation will depend on the high voltage.) In general, emission spectra are skewed to the red so that blue-green dyes appear in orange-red detectors rather than visa versa. The voltages (in millivolts, mv) at each OPAMP are shown in Table A-1 for single color compensation. Referring to Figure A-2 for green cells, FLB sees 20% of A, so that the signal at B is 1000 mv. Since B 20%A 1000 mv, OPAMP B has no output because the exact voltage has been subtracted from FLB and the output signal for OPAMP A is 5000 mv. Similarly for red cells, FLA detects only 5% of B so that the signal at A is 5% of 5000 mv or 250 mv. Since A 5%B 250 mv, the output of OPAMP A is zero and the compensated output signal from OPAMP B is 5000 mv. What happens if cells are double labeled with 5000 mv of green fluorescence and 5000 mv of red fluorescence at the same compensation settings shown above? The voltages at each OPAMP for doubly labeled cells are shown in Table A-2. The FLA detector has a signal voltage equal to green 5% of red or 5000 mv 250 mv or 5250 mv and the FLB detector has a signal voltage equal to red 20% of green or 5000 mv 1000 mv 6000 mv. The compensation (A - 5%B) produces a negative input voltage for OPAMP A of 300 mv (6000 0.05) and the negative input for (B-20%A) OPAMP B is 1050 mv (5250 0.2). This Table A-2 OPAMP Input Voltages (ma) for Double-Labeled Cells Input Output OPAMP A 5250 A 5% B 300 4950 OPAMP B 6000 B 20% A 1050 4950 produces an output from A of 4950 mv (5250 300). The output from OPAMP B is also 4950 mv (6000 1050). Thus, each fluorochrome has an output voltage that is somewhat less than when they were tested as single labels. This will not be noticeable when both antibodies produce very bright fluorescence, but will produce an increasingly noticeable effect as under-compensation when fluorochromes have increasingly disparate intensities. This can be easily demonstrated by substituting different voltage values for the FLA and FLB voltages and computing the output voltage for each OPAMP as shown in Table 2. As we increase the number of fluorochromes with overlapping spectra and their brightness, the voltage differences for double, triple and quadruple labeled cells become more complex and exacerbated, producing increasing divergence between single and multiple labeled cells. This becomes increasingly noticeable and problematic. It is hoped that this brief description will be useful in understanding the effects of multiple fluorescent signals on compensation. The degree to which these effects become noticeable will depend on the fluorochromes relative intensities on the cells being measured. By using appropriate compensation controls, the precise optimal settings for reliable measurement of fluorescence can be achieved. Questions are welcome and may be addressed to: stewart@sc3101.med.buffalo.edu. ACKNOWLEDGMENTS The authors wish to thank Drs. Michael Loken and Gerald Marti for their review and contributions to the appendix.