Natural killer (NK) and NK T (NKT) cells are important in innate immune defense. Their unequivocal identification requires at least four antigens. Based on the expression of additional antigens, they can be further divided into functional subsets. For more accurate immunophenotyping and to describe multiple expression patterns of leukocyte subsets, an increased number of measurable colors is necessary. To take advantage of the technologic features offered by slide-based cytometry, repeated analysis was combined with sequential optical-filter changing.
Human peripheral blood leukocytes from healthy adult volunteers were labeled with antibodies by direct or indirect staining. Tandem dyes of Cy7 (phycoerythrin [PE]-/allophycocyanin [APC]-Cy7), Cy5.5 (PE-/APC-Cy5.5), and PE-Cy5 and the fluorochromes fluorescein isothiocyanate (FITC), PE, and APC were tested alone and in combinations. Optical filters of the laser scanning cytometer were 555 DRLP/BP 530/30 nm for photomultiplier tube (PMT) 1/FITC, 605 DRLP/BP 580/30 nm for PMT 2/PE, 740 DCXR/BP 670/20 nm for PMT 3/Cy5/APC, and BP 810/90 nm for PMT 4/Cy7. Filter PMT 3 was replaced for detection of PE/Cy5.5 and APC/Cy5.5 by 740 LP/BP 710/20 nm and the sample was remeasured. Both data files were merged into one to combine the different information on a single-cell basis. The combination of eight antibodies against CD3, CD4, CD8, CD14, CD16, CD19, CD45, and CD56 was used to characterize NK and NKT cells and their subsets.
In this way Cy5.5 is measurable at 488-nm and 633-nm excitation. Further, with the two different filters it is possible to distinguish Cy5 from Cy5.5 in the same detection channel (PMT 3). With this method we identified NK and NKT cells, subsets of NK (CD3−16+56+, CD3−16+56−, CD3−16−56+) and NKT (CD3+16+56+, CD3+16−56+) and their CD4+8−, CD4−8+, CD4−8− and CD4+8+ subsets.
Natural killer (NK) and NK T (NKT) cells are members of the innate immune system and are able to recognize and eliminate non-self cells without previous immunization (1). Human NK cells do not express CD3 but do express CD56 or CD16 or both (2, 3). Thus, they are different from NKT cells that express CD3 and CD56 (3). NKT cells express a conserved canonical T-cell receptor (TCR; in humans, mostly Vα24Jα18-Vβ11) that is thought to recognize a self-antigen mimicked by the glycolipid αGalCer (4). They can rapidly secrete large amounts ofT-helper type 1 and 2 cytokines on TCR ligation and may play an important role in immunoregulation (5). A subset of human peripheral blood NKT cells also expresses CD8α and CD8β, another subset CD4 (2). Because these different subsets of NK and NKT cells differ in their biological function such as their response toward chemokines (2) or cytolytic effector function (3, 6), their unequivocal identification is a prerequisite for a better understanding of their biological role. Their identification requires at least four antibodies, namely CD3, CD16, CD45, and CD56.
Since the introduction of immunophenotyping by flow cytometry, multicolor analysis became of increasing clinical and research importance. On the one hand, due to increasing knowledge of the complexity of the immune system, more subsets of leukocytes are characterized that can only be identified unequivocally by their expression pattern of a multitude of surface antigens. On the other hand, in patients with low blood volume such as critically ill neonates, every effort has to be made to decrease the needed blood volume. This task can be accomplished by further increasing the number of fluorochromes and, hence, the characteristics per cell acquired within one analysis. During recent years this demand has led to the development of polychromatic flow cytometry (7), where up to 11 or more colors can be analyzed in a single run. However, in many applications and specific clinical settings, analysis by flow cytometry seems impractical because it needs larger sample volumes and does not enable visual inspection for verification of the events' morphology. Under these circumstances, slide-based cytometry (SBC) offers an alternative for multiparametric analysis by using only minimal sample volumes and allowing direct visual control (8). This can be done by typical SBC instruments such as the laser scanning cytometer (LSC) (9–11), the scanning fluorescent microscope (12), or the confocal laser scanning microscope (13).
Recently, we demonstrated the feasibility of six-color immunophenotyping on the slide by using a commercial LSC, with slight filter setting modifications (14). To further increase the number of measurable colors by LSC, we have adapted the system for the detection of the tandem dyes of Cy5.5. Because the LSC provides space for only four photomultiplier tubes (PMTs), an increase of simultaneously measurable colors is not feasible without major changes of the instrument (8). Therefore, for simultaneous analysis of up to eight colors, a specific feature of the LSC was used: the merging of data files. Because events on the slide are immobilized and the x and y positions of each event are saved in the data file, the same specimen can be analyzed several times with different filter settings. The resulting .fcs data files can then be merged into one virtual composite data file. According to the x and y coordinates of the cells, a single combined (“merged”) .fcs file represents a multidimensional dataset for each measured single-cell event. This feature is unique to SBC because the fixed position of a cell on the slide provides the only stable parameter that allows specific allocation of different results of analyses to the same cell. This merge feature, for example, has been used by others to combine vital and nonvital data on cultured cells before and after fixation in studies of apoptosis (15, 16).
Polychromatic immunophenotyping is of particular importance for the characterization of specialized leukocyte subsets such as memory, antigen-specific, and regulatory T cells, dendritic cells, and NK cell subtypes. In pediatric cardiology and cardiovascular surgery, NK cells are of particular interest. Pediatric cardiovascular surgery with cardiopulmonary bypass (CPB) may cause excessive and harmful postoperative inflammatory syndrome that leads to capillary leakage or even organ dysfunction (17). An increase in the NK cell count in the peripheral blood is the most prominent cellular feature of pediatric cardiovascular surgery with CPB but is absent without CPB (our unpublished results). Further, a preoperatively increased NK cell count is predictive for adverse surgical outcome or severe late complication in patients with a univentricular heart, with protein-losing enteropathy (18).
To further characterize lymphocytes and, in particular, NK and NKT subsets, immunophenotyping by six or more colors is required. To this end we tested whether eight-color phenotyping by LSC was technically feasible. We developed and established an eight-color immunophenotyping panel for LSC analysis of peripheral blood leukocytes. Using this panel we could not only distinguish between NK and NKT cells and their subsets but also determine the differential blood picture from only 40 μl of peripheral blood. With this novel approach we could characterize more than 13 distinct lymphocyte populations and subdivide different subsets of NK and NKT cells. Further, it was possible to distinguish regular from proinflammatory monocytes. In this initial study we used blood samples from healthy adult volunteers for the test system. With this new assay NK and NKT cell subsets can be analyzed in pediatric patients with only a minimum required blood volume.
MATERIALS AND METHODS
The instrumentation and software of the LSC (CompuCyte Corporation, Cambridge, MA, USA) are explained in detail elsewhere (9–11, 14). In the setup for six-color immunophenotyping, the following bandpass filters are used: 530/DF30 (green, filter cube 1), 580/DF30 (orange, filter cube 2), HQ670/50M (red, filter cube 3), and a D810/90M filter (far red, filter cube 4). For eight-color immunophenotyping, filters for the detection of green, orange, and far-red dyes are left unchanged (Fig. 1). For phycoerythrin (PE)/Cy5 and allophycocyanin (APC) measurements, filter cube 3 was replaced by a 670/20-nm bandpass filter and a 740DCXR dichroic filter. For detection of tandem dyes PE-Cy5.5 and APC-Cy5.5, a 710/20-nm filter in combination with a 740LP filter is used in filter cube 3 (optical filter cubes for PE-Cy5/APC and Cy5.5 were purchased from AHF Analysetechnik AG, Tübingen, Germany).
Staining of human peripheral blood leukocytes was performed as described elsewhere (14). Blood from infection-free healthy adult volunteers (age range 24–45 years, three women and five men) that was anticoagulated with ethylenediaminetetraacetic acid was collected.
Single-color staining (spillover matrix).
To test crosstalk of the various fluorochromes, 40-μl aliquots of blood were mixed with 5 μl of monoclonal anti CD3-biotin antibody (Caltag Laboratories, Hamburg, Germany) and incubated for 20 min at room temperature in the dark. Then 5 μl of streptavidin (diluted 1:10 with phosphate buffered saline; Sigma Diagnostics, St. Louis, MO, USA) conjugated to different fluorochromes (fluorescein isothiocyanate [FITC], PE, PE-Cy5, PE-Cy5.5, APC, or APC-Cy5.5; Caltag Laboratories) was added, mixed, and incubated for an additional 20 min. In cases of PE-Cy7 (Beckman Coulter, Krefeld, Germany) and APC-Cy7 (Caltag Laboratories) with low relative brightnesses (14), 2.5 μl of undiluted conjugate was added.
Two-color staining to test the filter-change system.
Forty microliters of ethylenediaminetetraacetic acid blood was mixed with two antibodies (5 μl each) labeled with CD16:PE-Cy5 (clone 3G8) and CD19:PE-Cy5.5 (SJ25-C1) or CD3:APC (S4.1) and CD8:APC-Cy5.5 (3B5; all antibodies from Caltag Laboratories), respectively. As a control, one antibody (labeled with Cy5 or Cy5.5) was used.
For multicolor staining panels, directly labeled antibodies against surface antigens were added at the first staining step. The following fluorochromes and antibodies were used: FITC:CD14 (clone MøP9, recognizes lipopolysaccharide receptor on monocytes, macrophages, and granulocytes; BD Biosciences, San Jose, CA, USA), PE:CD4 (SK3, CD4 antigen on T-helper cells and monocytes; BD Biosciences), PE-Cy5:CD56 (B159, isoform of neural cell adhesion molecule (N-Cam); BD Biosciences), PE-Cy5.5:CD19 (SJ25-C1, B-cell differentiation antigen; Caltag Laboratories), PE-Cy7:CD16 (3G8, FcγRIII on neutrophils, monocytes, NK cells; Beckman Coulter), APC:CD45 (HI30, leukocyte common antigen; Caltag Laboratories), APC-Cy5.5:CD8 (3B5, α chain and αβ heterodimer of CD8 antigen on cytotoxic T cells; Caltag Laboratories), and APC-Cy7:CD3 (S4.1, ϵ chain of the CD3 antigen/TCR complex; Caltag Laboratories). Optimal antibody concentrations were tested by titration. After antibody staining, red blood cells were lysed for 12 min with 1 ml of FACS Lysing Solution (BD Biosciences) at room temperature, specimens were washed twice in phosphate buffered saline, placed onto a conventional glass microscope slide (Menzel Gläser, Braunschweig, Germany), mounted with 20 μ of Fluorescent Mounting Medium (DAKOCytomation, Glostrup, Denmark), and analyzed after brief storage at 4°C in the dark to minimize cell movement. Initial experiments showed better resolution of fluorochromes (especially of FITC) when samples were covered with Fluorescent Mounting Medium compared with specimens mounted with glycerol/buffered saline (75%/25%, v/v) used in our previous studies (14) (Fig. 2).
Sample Analysis and Data Display
The principle of sample analysis by LSC and appropriate instrument settings were reviewed by two authors (A.T. and A.G.) (8). Analysis was performed with the 20× objective. Triggering was set on the forward scatter (FSC) signal (offset 2,092, gain 18, minimal area 8 μm2, threshold 5,000, add pixel 10; for more details on instrument settings, see Gerstner et al. ). To guarantee complete detection of all cells, these settings were modified depending on cell density on the slide, time after preparation, and other physical parameters that influence cell morphology and light scatter. Settings for the PMTs were as follows: gain 255, offset 2,075 except for APC (2,073) and APC-Cy7 (2,095). PMT voltages were optimized by several tests. In all presented experiments, PMT voltages were 35% for FITC, 24% for PE, 40% for PE-Cy5 and PE-Cy5.5, 55% for PE-Cy7, 45% for APC and APC-Cy5.5, and 60% for APC-Cy7. The following data for 20,000 to 40,000 cells were acquired: area, perimeter, x position, y position, and fluorescence integral and MaxPixel in all channels.
Measurements were performed in two steps. At the first measurement, the “Cy5” filter cube (A; Fig. 1) in position 3 was used, and FITC, PE, PE-Cy5, PE-Cy7, APC, and APC-Cy7 fluorescences were recorded. Then the “Cy5” filter was replaced by the “Cy5.5” filter (B). A second measurement was performed where data of PE-Cy5.5 and APC-Cy5.5 were acquired. These two data files were then merged into one file by using the “merge” function of WinCyte software (CompuCyte; match distance 8 μm, exclude distance 8 μm). After merging, we checked that the merge worked properly, i.e., that cells did not move between measurements (Fig. 3). In case of improper merge (merged events < 80% and Pearson's coefficient of correlation < 0.95), the analysis was discarded. Data for FITC, PE, PE-Cy7, and APC-Cy7 were taken from the first measurement to decrease the effects of photobleaching (8). In the case of APC-Cy7, the first filter set also yielded higher resolution parameter and relative brightness (P < 0.03, t test; Table 1).
Table 1. Fluorescence Resolution and Relative Brightness of the Eight Fluorochromes*
Human peripheral blood leukocytes were stained with saturating concentrations of CD3-biotin followed by streptavidin conjugated to one of the listed fluorochromes and analyzed by the LSC with different filters in filter cube 3 (A or B; Fig. 1). Data were gated on lymphocytes (for example, see Fig. 4.1). Results show the mean and the range of four representative experiments.
The mean fluorescence of CD3+ lymphocytes minus mean fluorescence of CD3− lymphocytes was divided by the sum of the respective standard deviations of CD3+ and CD3− lymphocytes according to Wood (28), with minor modifications.
Median fluorescence of CD3+ lymphocytes divided by the fluorescence of CD3− lymphocytes on the same sample. Values were normalized to the relative brightness of FITC (7).
For all analyses, the following gating strategy was used. First, debris and artifacts were excluded based on their low FSC MaxPixel and cell area values (Fig. 4-1a). For the data shown in Figure 4-2, residual signals were also gated on lymphocytes based on their relatively low FSC integral and cell area values (Fig. 4-1b). Figure 4-1b shows a clear discrimination between lymphocytes (Fig. 4-1c) and residual cells such as neutrophils and monocytes (Fig. 4-1d). Only these lymphocyte data are shown in Figure 4-2. Histograms of all colors were then generated without further gating (Figs. 4 and 5). Fluorescence resolution and fluorescence spillover (Tables 1 and 2) were calculated from data of at least four single-color experiments (typical example is shown in Fig. 4-2; further details are provided in notes to Tables 1 and 2). Data of eight-color panels, developed for lymphocyte subtyping, were split into leukocyte subsets by a gating cascade (Fig. 6). Cells from every subpopulation were directly visualized in the LSC by relocalization to confirm regular morphology (not shown).
Table 2. Spillover Matrix for Eight-Color Laser Scanning Cytometry*
First laser (488 nm)
Second laser (633 nm)
Ar red (A)
Ar red (B)
Ar long red (A)
HeNe red (A)
HeNe red (B)
HeNe long red (A)
Each value is a percentage of an integral fluorescence signal measured in each detector (normalized to 100% for the fluorescence's primary detector). For optimal detection, of Cy5 filter A and for Cy5.5 filter B was used in filter cube 3. Each row is related to the emission spectrum of the fluorochrome named in the first column given the particular combination of lasers, filters, detectors, and PMT-settings used in our instrument. For instance, the measured integral fluorescence of FITC in the PE-channel Ar orange is 4.26% of that in the FITC channel Ar green. Data were calculated using the median fluorescence integral signal of CD3+ lymphocytes with background (i.e., median integral fluorescence of CD3− lymphocytes from the respective channels) subtraction.
Selection of Antibodies and Dyes for Cell Identification
Based on the spillover matrices, it is clear that for optimal data display and analysis cross-compensation of all colors would be useful. At present, cross compensation for all colors is not possible with WinCyte software. Only in a single dot plot and histogram can compensation of one color be performed (e.g., we compensated the spillover of the PE fluorescence into the CD56:PE-Cy5 histograms). To unequivocally identify different leukocyte subpopulations with the eight-color matrix, it was therefore important to combine the antibodies and the respective dye labels in a way that each cell subset could clearly be identified by logical gating cascades. Figure 6 shows that, by the experimentally determined dye and antibody combination, all eight colors can be resolved and immunophenotyping can be well performed (for more details, see Results). Because of negative control for correct positioning of gates for CD56+ cells in the eight-color panel, a control sample stained without CD56 antibody was used (data not shown).
Separate Analysis of Eight Fluorochromes
The filter sets (Fig. 1) were suitable for detecting the emission of all eight fluorochromes used in this study. By using the anti-CD3/biotin/streptavidin system, positive versus negative lymphocytes were well separated with all fluorochromes (Fig. 4).
Staining qualities were compared by calculating the resolution parameter and the relative brightness (Table 1). Surprisingly, the resolution parameter for PE-Cy7 (filter A) was significantly higher than for all other dyes (P < 0.02, paired two-tailed Student's t test; except for FITC, which was not significant). The resolution parameter for FITC was higher than that for all residual colors (P < 0.02) except APC. Based on the resolution parameter, filter A in position 3 was more suitable to resolve the Cy7 tandem fluorescences than was filter B (P < 0.03). Resolution parameters for PE-Cy5 and PE-Cy5.5 and for APC and APC-Cy5.5 measured with respective filters (A for APC/Cy5 or B for Cy5.5) were not significantly different. This yields the following list of order for the resolution parameters: PE-Cy7 (A) ≥ FITC > PE > PE-Cy5 ≥ PE-Cy5.5 > APC ≥ APC-Cy5.5 > APC-Cy7
Second, the relative brightness (i.e., median fluorescence integral ratio of stained to unstained cells) was calculated as a parameter of the separation of both populations and is shown relative to the value obtained with FITC (6). As for the resolution parameter and relative brightness, FITC had significantly higher values than most of the other fluorochromes (P < 0.01, except for PE and PE-Cy7). The relative brightness of the APC-Cy7 conjugate was higher with filter set A in position 3 than with filter set B (P = 0.001). The highest relative brightness was obtained for PE and decreased in the following list of order: PE ≥ FITC > PE-Cy7 (A) > APC > PE-Cy5 > APC-Cy7 (A) > PE-Cy5.5 ≥ APC-Cy5.5
Thus, both Cy5.5 tandems showed the lowest relative brightness followed by APC-Cy7. The brightness of tandem dyes depends on the stability of the link between the donor dye and the acceptor dye. Excessive exposure to light decreases resolution and increases undesired fluorescence of the donor dye, PE. In contrast to our previous experiments (14), FITC had a very good resolution parameter and relative brightness near that of PE. This improvement resulted from the antifading mounting agent (Fluorescent Mounting Medium, DAKOCytomation) that probably decreased its susceptibility to photobleaching and stabilized the pH. We compared the effect of the Fluorescent Mounting Medium with that of glycerol/buffered saline that was used in our previous experiments (14). The most prominent difference was found for FITC, with about a four times higher resolution parameter and relative brightness in Fluorescent Mounting Medium; PE- and APC-labeled cells showed about 25% higher relative brightness (Fig. 2).
Nevertheless, in the present study, we determined a lower resolution parameter and relative brightness for APC (Tables 1 and 2) compared with the six-color protocol (14). This is due to the bandpass filter in filter cube 3. The bandpass of the new “Cy5” filter (A) is narrower (670/20 nm instead of 670/50 nm) so that a part of the APC emission is not measurable (Fig. 7). The resolution parameter of PE-Cy5 is increased but that of APC is decreased because the 670/20 nm filter is optimized for PE-Cy5 emission maximum. This was necessary because of the two additional fluorochromes used (PE-Cy5.5 and APC-Cy5.5). With the narrower bandpass of the filters for PE-Cy5 and APC, we were able to decrease the spillover of the Cy5.5 tandems into this channel. The higher relative brightness of PE-Cy7 and APC-Cy7 was achieved by using higher concentrations of antibodies or streptavidin-fluorochrome conjugates.
All fluorochromes showed detectable and in some cases substantial signals in more than one detector (Fig. 4-2 and Table 2). Most spillover signals arose in the PE channel (Ar 2; spillover from FITC, PE-Cy5, PE-Cy5.5, and PE-Cy7) and in the APC channel (HeNe 3; spillover from PE-Cy5, PE-Cy5.5, APC-Cy5.5, and APC-Cy7). In the PE-Cy5.5 channel (Ar 3, “Cy5.5” filter), there was only minimal spillover from one fluorochrome (PE-Cy5). There was no substantial difference in spillover into Ar 4 and HeNe 4 irrespective of whether filter A or B was used in filter cube 3 (Table 2). However, because resolution parameters of PE-Cy7 and APC-Cy7 were higher with filter A in filter cube 3 (Table 1), these data were taken for further analysis.
Four-Color Measurement in One Channel
Discrimination of cells labeled with PE-Cy5 and PE-Cy5.5 or APC and APC-Cy5.5 was feasible. For discrimination of PE-Cy5 and PE-Cy5.5 or APC and APC-Cy5.5, a measurement with filter set A (“Cy5”) in filter cube 3 followed by a second measurement using filter set B (“Cy5.5”) was necessary. The two obtained data files were then merged into one file where all fluorescences could be displayed as separate colors (Fig. 5). Although the fluorescences of PE-Cy5 and APC produced only minimal spillover into the Cy5.5 detector, the Cy5.5 conjugates generated a prominent signal in the Cy5 channel. Thus, the Cy5.5 fluorescences were located near a diagonal when the Cy5 vs. Cy5.5 channel was displayed. The fluorescences of APC and PE-Cy5 are located in the dot plot more in the Cy5 channel direction due to the lower spillover into the Cy5.5 channel.
Eight-color combination of antibodies and fluorochromes for NK and NKT cell characterization had to be chosen carefully to overcome the lack of cross compensation in the WinCyte software and to unequivocally identify each subset. In initial tests single antibodies and antibody cocktails conjugated to different fluorochromes were tested. This was necessary to find combinations where no cross compensation was required and the resolution of each leukocyte subset was sufficient. These combinations were CD14: FITC / CD4: PE / CD56: PE-Cy5 / CD19: PE-Cy5.5 / CD16: PE-Cy7 / CD45: APC / CD8: APC-Cy5.5 / CD3: APC-Cy7. This panel allowed identification of subsets by appropriate gating strategies and to avoid erroneous cell gating due to fluorescence spillover in two-parameter dot plots. To this end, fluorochromes with most prominent spillovers (e.g., FITC into PE channel) were used to label cell subpopulations where high expression on both cells types was not expected. Leukocytes could thereby be differentiated due to their different expression levels of CD45 (APC) and CD14 (FITC; Fig. 6, top right dot plot). Thus, monocytes (CD45dimCD14bright) and neutrophils (CD45dimCD14dim) could be excluded from analysis of lymphocytes (CD45brightCD14−). Therefore, NK and NKT cells that might have been CD16high were unequivocally identified because neutrophils that were also CD16high were eliminated from the lymphocyte gate. In addition, fluorescence spillover of FITC (CD14) into the PE channel (CD4) was irrelevant for the lymphocyte population. The residual fluorochromes that would produce substantial spillover into the lymphocyte gate (PE-Cy5,PE-Cy5.5, APC-Cy5.5, and APC-Cy7) were used to tag antibodies that detected exclusively lymphocyte antigens (CD56, CD19, CD8, and CD3, respectively). In addition, antibodies essential for subtyping of NK and NKT cells (CD4 and CD8) were combined with fluorochromes where no spillover was detectable (PE and APC-Cy5.5). With this combination double positive cells could not be the result of spillover.
Leukocyte, NK, and NKT cell subtypings by this eight-color panel are shown for a typical example in Figure 6. This panel was used to determine percentages of the respective leukocyte and lymphocyte populations in blood samples from eight healthy adult individuals (age range 24–45 years, three women and five men; details on leukocyte subsets and their sizes are presented in Table 3). For analysis, debris, artifacts, and aggregates were first excluded due to their FSC MaxPixel and area values (Fig. 6, top left). Correct gating was verified by rescanning the gated cells and morphologic evaluation. Discrimination of monocytes, lymphocytes, granulocytes, and basophils was possible by their expression levels of CD14 and CD45 (top right dot plot). The lack of CD14 on basophils allowed their separation from neutrophils and eosinophils (CD14dim) and monocytes (CD14bright). Within the monocyte subset we found a minor but prominent population of CD16+ cells with slightly decreased CD14 expression. Based on the work of Belge et al. (19), these cells were identified as proinflammatory monocytes. The CD16dim cells within the granulocyte gate were eosinophils.
Table 3. Percentage and Cell Size of Leukocyte, Lymphocyte and NKT and NK Subsets Determined by Eight-Color Immunophenotyping*
Mean of eight healthy adult individuals. The respective mean cell areas are normalized to that of neutrophils. Values were determined from peripheral blood of eight healthy adult volunteers. − negative; +/−, expressed on subsets; + to +++, dim to bright; NK, natural killer, DN, double negative; DP, double positive; Min – Max, minimum to maximum;
The lymphocytes could be further subdivided into four subsets: CD3+CD16−, CD3+CD16+, CD3−CD16−, and CD3−CD16+ (Fig. 6, second row, third picture). B cells in the CD3−CD16− population were identified by their CD19 expression (third row). They were negative for CD56 (not shown). The other three lymphocyte subsets were CD19− (not shown). The majority (93%) of CD3+CD16− lymphocytes were CD56− T cells (fourth row, first histogram) with subsets of T-helper cells (CD4+CD8−), T cytotoxic cells (CD4−CD8+), and CD4−CD8− (double negative [DN]) T cells (fifth row, first histogram). A minor population of these CD3+CD16− cells was CD56+ (NKT, 7% of CD3+16− cells). These NKT cells could be further subdivided into CD4+ (19% of this NKT cell population), CD8+ (69%), DN (9%), and CD4+CD8+ (double positive [DP]; 3%; sixth row, first histogram). The CD3+CD16+ population (fourth row, last histogram) also consisted of NKT cells (CD3+CD16+CD56+, 2% of all CD3+ cells) and activated T cells (CD3+CD16+CD56−) (2%). The CD3+CD16+CD56− subset was mostly CD4+ (48%) or CD8+ (23%; fifth row, last histogram). Most CD3+CD16+CD56+ NKT cells were CD8+ (45%) or DN (42%; sixth row, last histogram). The NK cells (fourth row, second and third histograms) could be divided into phenotypes CD3−CD16−CD56+ (15% of all NK cells), CD3−CD16+CD56− (30%), and CD3−CD16+CD56+ (55%). They were mostly DN (66–76%). An unknown residual cell population within the lymphocyte gate was found that expressed a CD3−CD16−CD19−CD56− phenotype and contained a minor CD4+ subset.
To validate the correct identification of the different leukocyte subsets, we determined the respective mean cell areas (normalized to that of neutrophils; Table 3). As expected, neutrophils and eosinophils were largest cells followed by monocytes; the smallest cells were basophils and lymphocytes. The respective areas of T cells and NK cells were in a similar range. Cells expressing CD16 (e.g., proinflammatory monocytes or activated T cells) had a larger cell area than their CD16− counterparts. Also, CD16+ NK and NKT cells were about 10–20% larger than CD16− NK or NKT cells. In addition, cells were relocalized and optically inspected in the LSC to check their morphology. This analysis confirmed that the DP NKT cells were single cells that coexpressed CD4 and CD8. These cells with the phenotype CD45brightCD14−CD19−CD3+CD16−CD56+CD4+CD8+ have not been reported previously.
The increasing knowledge about the diversity of phenotype and function of cells of the immune system demands that a larger amount of colors is measurable simultaneously. We previously demonstrated that immunophenotyping with six different fluorochromes by LSC is possible (14). This number of fluorochromes remains smaller than what can be achieved by newer flow cytometry systems (7). Nevertheless, SBC has some specific advantages and expands the opportunity to measure even more than eight colors. An important advantage of SBC is that cells that have been analyzed are not lost and may be used for additional analyses and morphologic judgement (8) or subsequent analysis of the same specimen with different instrument configurations. Although our study was performed on one type of SBC instrument, the LSC, the dye combination and the approach of filter changes is in principle applicable for other types of SBC instruments such as the scanning fluorescent microscope (20), spectral imaging (13), and scanning cytometry (21).
With the combination of filter change, remeasurement, and merging feature, we are able to combine fluorochromes with relatively similar emission spectra (e.g., PE-Cy5 and PE-Cy5.5). A problem for data analysis is the spectral overlap of some colors into channels where other fluorochromes are measured. To avoid these false-positive events, the extent of spectral overlap has to be quantified and included into the analysis. With WinCyte software only one false-positive signal per channel can be subtracted, i.e., cross compensation is not possible. If cross compensation is essential, data can be exported and analyzed by alternative software (e.g., Summit, DakoCytomation). However, this requires the export of the data from the FCS format of the LSC into FCS2.0 that is generally readable by various flow cytometric softwares. With the current version of WinCyte, this is difficult to perform in a standardized way.
Therefore, antibody and dye combinations have to be tested carefully to avoid data that can be misinterpreted. We recommend the use of near infrared dyes for the unequivocal identification of subsets. Dyes with substantial spillover (e.g., FITC and PE) should be used as labels for separate lineages. Antigens that are expected to be coexpressed on the same cell should be selected so that they have only minimal spillover.
With the present eight-color panel we are able to characterize different leukocyte populations, in particular the different subsets of NK and NKT cells. At least eight different antigens are necessary to separate lymphocytes from the residual leukocyte populations and to subdivide NK and NKT cells. We found three different types of NK cells characterized by the CD3−CD16−CD56+, CD3−CD16+CD56−, and CD3−CD16+CD56+ expression patterns. Suzui et al. (3) reported two phenotypically and functionally distinct NK subsets in humans using three-color flow cytometry. Their subsets may overlap with our three subsets but are probably not completely identical. This is most probably due to the greater brightness of CD56 achieved by these investigators (3). Their major NK population was CD16brightCD56dim (7.9% of all lymphocytes) and may resemble to our CD3−CD16+CD56+ subset (7.2%; Table 3). However, their CD16dim/−CD56bright subset (0.4%) may agree with our CD3−CD16−CD56+ (1.9%). The first subset has a higher cytotoxicity, a lower expression of CD44 and CD62L, but higher expressions of CD11a and CD18 (3). These investigators also observed a subset with a CD3−CD16+CD56− expression pattern but did not comment on their frequency and function. These different types of NK cells seem to represent different stages of activation (22). Loza et al. (22) found that CD56+lowCD16+ cells can become CD56+highCD16− cells by down-modulation of CD16 after culture with interleukin-15 and interleukin-12. That implies that CD16+CD56− NK cells, found with this panel, are the first stage of NK cell activation. We could further subdivide the NK subsets by their expression of CD4 or CD8 and found that most expressed neither CD4 nor CD8.
Using our approach we could detect NKT cells by the coexpression of CD3 and CD56. With the eight-color panel we found that 80% of the NKT cells were CD16−CD56+ and about 20% were CD16+CD56+. The latter population may resemble the CD16+ subset within the CD56+NKT cells reported by Ou et al. (23) that comprised fewer than 5% of all cells in a β-cell antigen-specific T-cell line. The two NKT cell populations (CD3+CD16−CD56+ and CD3+CD16+CD56+) can be further subdivided based on their CD4 and CD8 expressions. Within the CD3+CD16−CD56+ cells we found a prominent population of CD8+ cells (69%) and a minor cell group expressing CD4 (19%). In contrast, 42% of the CD3+CD16+CD56+ NKT cells were negative for CD4 and CD8 and 45% for CD8+. In mice CD1d-independent can be differentiated from CD1d-restricted NKT cells. A subset of both populations was DN for CD4 and CD8. The other subset of CD1d-independent NKT cells expressed only CD8, and the residual part of the CD1d-restricted NKT cells are CD4+ (24). Human CD1d-restricted invariant T cells express a semi-invariant TCR chain with a Vα24JαQ rearrangement, which preferentially pairs with a restricted set of TCR β chains, including Vβ11 (25). These CD1d-restricted T cells (NKT) play a role in immunoregulation and are of special interest as a target of drug development in autoimmune diseases (26). Further, due to their functional involvement of self ligands during infection, they are declared as a “bridging system” between innate and acquired immunities (27). Similar to CD56+highCD16− NK cells, the CD56+ NKT cells also might represent the last stage of activation (23), thus indicating that NK and NKT cells use the same pathway of activation.
To the best of our knowledge, this is the first publication that reports different subsets of NK and NKT cells concerning their expression of CD3, CD4, CD8, CD16, and CD56. In particular, we report some minor DP NK and NKT subsets. These subpopulations could be the subject of further characterization with respect to phenotype and function in future studies. This study demonstrates the feasibility of eight-color immunophenotyping on LSC and shall act as an initial study to characterize NK and NKT subsets. In further studies the cell populations described in this report could be confirmed by analysis of more individuals and examined in patients with specific diseases (e.g., characteristics in congenital heart disease, trauma such as cardiovascular surgery with CPB).