The authors acknowledge Geert Weijers, Ruud Huisman from Beckman Coulter, The Netherlands, and Mario Koksch, Beckman Coulter Europe for support with materials and reagents and for helpful discussions. They thank Maria Jesus Capel and Isabel Mainar for microscopic differentiations.
Differential white blood cell count (dWBC) is an important diagnostic tool for adequate treatment, in particular for patients with a hematological malignancy, infection, or inflammation. In general, automated hematology analyzers cope well with the bulk of samples. Aberrant samples are flagged for microscopic review based on pre-set criteria. Microscopic differentiation requires preparation and staining of a blood smear and manual counting by a skilled technician. This technique has several well known disadvantages. Because of the low number of cells counted (100) and uneven distribution of nucleated cells on the slide, statistical sample errors are significant (1, 2). Interobserver subjectivity, even between experienced technicians within the same laboratory, is another known problem. The time required for training and maintaining sufficient expertise among technicians to allow round the clock service is significant. Reporting dWBC results from complicated patient samples on time for adequate treatment can be a logistic challenge.
Studies on the expression of surface antigens on leukocytes enabled unambiguous classification of leukocytes far beyond the possibilities of microscopy. Flow cytometry is a suggested reference method for dWBC due to the immunological definitions of cell populations and the large number of cells counted (3). There are numerous reports on flow cytometric characterization of subsets of leukocytes in peripheral blood samples (4–10). In contrast, reports of a more complete leukocyte differential count by flow cytometry are limited (11–14). Differences exist in the antibody cocktails used, the leukocyte populations identified, and the classification methods.
Our aim was to develop a new 5-color flow cytometric assay, called Leukoflow, to replace and improve the microscopic dWBC in 24 h/7 days laboratory service. We designed a cocktail of antibodies combined with gating strategy and flow cytometric protocol allowing easy identification of 13 different leukocyte populations. For quantification of normoblasts a second flow cytometric staining was designed. We present the analytical validation of this assay, comparing Leukoflow with the automatic cell counter and microscopy.
MATERIALS AND METHODS
Samples and Routine Analysis
Blood samples, drawn by sterile venipuncture in tubes containing 5.4 mg K2EDTA (Becton Dickinson, #368856, Plymouth UK) from human patients undergoing routine diagnostic blood analysis between January 2009 and August 2010 in our institute, were analyzed by our automated hematology analyzer LH750 (Beckman Coulter, Miami, FL). We used material remaining after routine analysis, results were not reported to patients, so informed consent was not required. The patient groups are displayed in Table 1. The results of the LH750 5-part differentiation were used to classify samples as normal or abnormal according to the H20-A2-criteria of the Clinical and Laboratory Standards Institute (3). We did not use the flags generated by the LH750 as criteria for selection, both flagged and unflagged samples were tested. In total, 100 normal and 102 abnormal blood samples were included. All samples arrived in the lab approximately within 1 h after venipuncture and were processed within 4 h upon arrival. The bone marrow aspirate was collected in an EDTA-tube, passed through a 70 μm Nylon filter (BD Falcon, #352350, Erembodegem, Belgium) and processed similar as the blood samples.
Table 1. Distribution of pathological groups (one patient may be tested multiple times on different days)
Of each sample, blood smears were prepared manually, stained with May-Grünwald Giemsa dye (15) and sent to an external laboratory for review. Each of two qualified independent reviewers, unaware of the other results, manually differentiated 200 cells per blood smear, the average was taken for analysis.
Immunological definitions used to discriminate the different cell types are shown in Table 2. The allocation of antibodies to the channels of the flowcytometer and the clones (mouse monoclonals) and dilutions used are shown in Table 3. Antibodies were titrated for optimal detection of the target populations. Twenty microliter cocktail contains 2 μL of the titrated dilution of each antibody. Samples were diluted with PBS+0.5% BSA to obtain a WBC count <10•109 cells/L. Blood (20 μL) was stained with 20 μL antibody cocktail for 15 min in the dark. Next, 500 μL lysis solution (1.5 M ammonium chloride, 100 mM potassium hydrogen carbonate, 0.82 mM EDTA, pH 7.4) was added followed by 15 min incubation in the dark. Cells were centrifuged at 450G for 5 min, 500 μL supernatant was carefully replaced by 500 μL PBS+0.5% BSA. Immediate before analysis 20 μL Flowcount beads (Beckman Coulter, #7547053) were added. Samples were measured on an unmodified single laser FC500 flowcytometer (Beckman Coulter), for 5 min or until a maximum of 100,000 events. Spectral compensation was performed using Advanced Digital Compensation (ADC, Beckman Coulter). No data transformation of the raw measurements was performed. All samples were measured with the “Baseline offset” switched off. Before each use, the optics and settings of the flowcytometer were checked with Flow-Check and Flow-Set beads (Beckman Coulter, #6605359, #6607007, #737664).
Table 2. Immunological definition of cell types
CD138 strong pos
Neu Pre I
Draq 5 pos
Table 3. Allocation of markers to flowcytometer channels, antibody information, and expression profiles on blood cells
Catalog number (Beckman Coulter)
T-cell subset and monocytes
Neutrophils and NK-lymphocytes
Hematopoietic precursor cells/blasts
T-cell subset and NK-lymphocytes
Pre B-cells and plasma cells
In most channels two or three antibodies are combined and the expression is referred to as the combination of the markers. To identify all populations we developed a multistep sequential gating strategy (Fig. 1A and Table 4). Using different combinations of regions, cell populations are defined by orientating gates and selection regions as described in Table 4 and the results section. The legend gives the percentages and the absolute cell numbers in cells/μL of the populations.
Table 4. Sequential gating strategy
Number of pre-defining plot and orienting gate
Sequential gating strategy used to identify the cell populations. In the column “pre-defining plot and orienting gate” the plot in which the cell type is defined is given (bold), followed by the first selection for that population. Column “selection region” gives the region used in the pre-defining plot to select the cell population of interest. The final percentage of the cell population of interest is calculated by using a final gate, this is not a physical region visible in a plot, but is a definition. Note all plots, regions and gates for defining the cell populations are set on [Tm and not (A)] criteria. For position of regions see Figure 1. All cell populations are finally expressed as percentage of the gate “Leuco,” which excludes of counting beads and debris.
Histogram plot with all events over time ungated
Tm Selected by including events with proper counting rate, excluding artefacts
2) Histogram FL-5Tm and A
Cal and A
Cal and A Single beads gated by “Cal” for quantification. All beads excluded from additional analysis by region “A”
3) SS vs CD45/138 and4) CD34/16/56 vs CD45/138
Tm AND NOT(A) AND NOT (B1 AND Debris) AND NOT (B3 AND Debris) AND NOT (B4 AND Debris) AND (B1 OR B2 OR (B3 AND Blast))
4) CD34/16/56 vs CD45/138:Tm AND NOT (A) AND (B1 or B3)
Tm AND NOT (A) AND (B1 or B3 or B4) AND Debris
3) SS vs CD45/138:Tm AND NOT (A) AND (B1 or B2)
Tm AND NOT (A) AND (B1 or B2) AND Plasma cells
4) CD34/16/56 vs CD45/138:Tm AND NOT (A) AND (B1 or B3)
Tm AND NOT (A) AND (B1 or B3) AND Blast
8) SS vs CD14/4:Tm AND NOT (A) AND (B1 or B2) AND NOT (Plasma cells OR T +34)
Tm AND NOT (A) AND (B1 OR B2) AND NOT (Plasma cells OR T +34) AND M
10) CD34/16/56 vs CD3/71:Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO)
Neu Pre I
Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO) AND Neu Pre I
10) CD34/16/56 vs CD3/71:Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO)
Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO) AND Neu Imm
10) CD34/16/56 vs CD3/71:Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO)
Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M OR EO) AND Neu Mat
5) CD3/71 vs CD45/138:Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR B +34 OR Blast)
Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR B +34 OR Blast) AND T +34
11) CD3/71 vs CD14/4:Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR B +34 OR Blast) AND T +34
Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR B +34 OR Blast) AND T +34 AND T4
7) SS vs CD19:Tm AND NOT (A) AND (B1 or B2) AND NOT (Plasma cells OR Blast)
Tm AND NOT (A) AND (B1 or B2) AND NOT (Plasma cells OR Blast) AND B +34
6) SS vs CD34/16/56:Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR M OR T +34 OR Blast)
Tm AND NOT (A) AND (B1) AND NOT (Plasma cells OR M OR T +34 OR Blast) AND NK +34
9) CD34/16/56 vs CD45/138:Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M)
Tm AND NOT (A) AND B2 AND NOT (Plasma cells OR M) AND EO
Not defined in plot (Subtraction of cell populations in B1)
Tm AND NOT (A) AND B1 AND NOT (Plasma cells OR M OR T+34 OR B+34 OR NK+34)
Detection of Normoblasts
To detect and quantify normoblasts a second flow cytometric staining is used when the LH750 indicates presence of normoblasts. Twenty microliter of EDTA blood is stained with 2.5 μL of CD45-FITC (Immunotech, Marseille, France, #0647, clone ALB12, 10× diluted) for 15 min in the dark. Next, 500 μL of PBS/BSA and 10 μL of DRAQ solution (50 μM) are added (Biostatus, #BOS-889-001-R200, Shepshed, UK). DRAQ5 is suited for chromatin visualization in living cells and DNA labeling of erythroblasts (16). After 15 min in the dark, the sample is measured (Fig. 1B).
The results from the different methods for dWBC are compared by orthogonal regression analysis (17). Reproducibility of cellular classification according to the cell counter (LH750) and Leukoflow, respectively, is described with the coefficient of variation (CV%). Values for all absolute cell counts measured in adult normal blood samples by Leukoflow are described with conventional descriptive statistics.
When designing of our antibody cocktail, the immunological markers required by the Netherlands Society of Quality Assurance in Medical Laboratory Diagnostics (SKML) for leukemia and lymphoma immunodiagnostics were used as guideline. Our gating strategy is new and allows easy positioning of gates and discrimination of cell populations. Since we use a 10-antibodies/5-colors approach, most channels contain multiple antibodies (Table 3). The discrimination between two cell populations both positive for the combination of the two markers in that channel is made by another parameter, e.g., a gate or side scatter. Cell populations are selected in three steps. First, a selection of events is displayed in a dot plot. Next, a region selects a population. The final cell population is gated by taking the cells in that region and excluding events in other regions. This sequential gating strategy is explained in detail below (Table 4). The different cell populations are expressed as a percentage of the total leukocyte population and absolute cells numbers are calculated using flowcount beads. Data acquisition starts by including the events over time (Tm), enabling exclusion of artificial events such as air bubbles (Fig. 1A, Plot 1). The CAL region (Plot 2) calculates absolute cell counts using flowcount beads. Next the signal from the flowcount beads (region A) is excluded, resulting in the selection “Tm and not A”, used for all plots. Leukoflow analysis of a normal blood sample is shown in Figure 1A.
Identification of Debris, Blasts, Leukocytes, and Plasma Cells
Plot 3 (Fig. 1A) is pre-selected to exclude events located in the combined B1 and Debris regions, the combined B3 and Debris regions and the combined B4 and Debris regions and includes all other events located in B1, B2, and the combined B3 and blast regions. Plot 4 is preselected on events with low SS (quadrant B1 or B3). All events that fall in the Debris region and that fall in quadrants B1 or B3 or B4 of Plot 3 are classified as debris. Blasts have low SS (preselected on quadrant B1 or B3 of Plot 3) and are selected by the Blast region as CD45 intermediate and CD34/16/56 positive. Blasts are depicted in red and this color is dominant. When blasts are either CD3 or CD19 positive they appear, respectively, in red in the T-cell region in Plot 5 or the B-cell region in Plot 7. When the blasts are negative for CD3 and CD19, as is the case for myeloid blasts, they appear in red in the NK-region in Plot 6 (illustrated in Figs. 3A–3D). Depicting the lineage of origin of blasts gives valuable diagnostic information.
Leukocytes are defined in Plots 3 and 4 (Fig. 1A), as all events that are present in region B1, B2 or in both the B3 and blast regions combined, minus events that are present in both region B1 and debris, or both region B3 and debris or both region B4 and debris. CD138 is used to detect plasma cells, titrated to give a strong signal. In Plot 3, the region for plasma cells includes all events that are strong positive in the CD45/138 channel.
Identification of Lymphocytes
T-cells are defined in Plot 5, preselected on low SS (quadrant B1 or B3) and excluding events in region B+34. T-cells are defined as low SS, CD45 positive (B1) and double positive in Plot 5, selected by region T+34. This region is called T+34 because T-cell blasts, if present, also fall in this region. To get mature T-cells out of the T+34 population, cells in the regions “plasma cells,” “B+34,” and “Blast” are subtracted. Further subdivision of CD4 positive and CD4 negative T-cells is made with Plot 11, preselected to include quadrant B1 and T+34 cells, excluding plasma cells and B+34 events. Quadrant T4 in Plot 11 selects CD3/71 and CD14/4 double positive cells and upon exclusion of plasma cells, B+34 cells and blasts, CD4 positive T-cells remain. For definition of NK cells, CD45 positive cells with low SS are preselected (quadrant B1), excluding events present in both B1 and debris regions combined or in the plasma cells or M or T+34 regions. In Plot 6, CD34/16/56 positive cells are selected in region “NK+34.” Upon subtraction of the blasts, the NK-cells remain. B-cells are preselected as CD45 positive (B1 or B2), excluding plasma cells. In Plot 7, region B+34 defines a population of CD19 positive cells with low SS as B-cells and blasts of the B-lineage. Upon exclusion of the blasts, the B-cells result.
Identification of Monocytes
Monocytes are identified in Plot 8, preselected to contain cells from B1 or B2 quadrants (CD45/138 pos), excluding events falling in the plasma cell and T+34 regions. Monocytes are identified by region M as CD14/4 positive with intermediate SS, excluding the events in T+34 and excluding plasma cells. We did not include this in our protocol but an additional plot, pre-selected on monocytes, can discriminate between CD16 positive (pro-inflammatory) and CD16 negative monocytes after subtraction of blasts and NK-cells (18, 19).
Identification of Granulocytes
Granulocytes are pre-selected as CD45 expressing cells with a high SS (B2), excluding the plasma cells and monocytes. Eosinophils are discriminated from the neutrophils by region Eo based on their lower expression for CD34/16/56 and their higher expression of CD45/138 (plot 9). Within the neutrophilic lineage, different stages of maturation are discriminated in quadrants in Plot 10, preselected on CD45 positive cells with high SS (B2), excluding events falling in the regions Plasma cells, M or Eo. Mature neutrophils (Neu Mat) are CD34/16/56 positive and negative for CD3/71. Immature neutrophils (Neu Imm) have low or no expression for CD34/16/56 and are CD3/71 negative. Neutrophilic precursors (Neu PreI) have low or no expression for CD34/16/56 and express CD3/71. Quadrants in Plot 10 were placed identical for all samples. Basophils are defined as the residue of the cells in the B1 region in Plot 3 (CD45/138 positive with low SS) after exclusion of events in regions T+34, B+34, NK+34, M and Plasma cells.
Detection of NRBC
Presence of normoblasts in a blood sample can interfere when determining the absolute leukocyte count on a hematology analyzer. The leukocyte count of Leukoflow is not influenced by normoblasts and gives the absolute leukocyte count adjusted for the normoblasts. To quantify the number of normoblasts per 100 leukocytes a second flow cytometric staining was designed. For this, blood is measured without lysis and DRAQ5 is used for live gating, including all nucleated cells and excluding mature erythrocytes (Fig. 1B, Plot 12). Normoblasts are identified as DRAQ5 pos, CD45 negative cells in region NRBC (plot 13).
Absolute Leukocyte Count by Automatic Cell Counter and Leukoflow
The number of total WBC events collected during Leukoflow analysis ranges from 10,317 to 100,228 (median 77,943 and 74,211 events for normal and abnormal samples, respectively). Comparing the absolute WBC between the automatic cell counter and Leukoflow we found excellent correlations for normal [regression coefficient (r) = 0.98; orthogonal regression line y = 0.988x−0.216] and abnormal samples [r = 0.99; orthogonal regression y = 1.015x−0.433]. This demonstrates Leukoflow gives accurate absolute leucocyte counts.
Comparison of Automatic Cell Counter, Microscopy, and Leukoflow on Normal Samples
We compared the three methods for performing the 5-part leukocyte differentiation on 100 normal blood samples (Table 5). The results from Leukoflow and automatic cell counter correlate very well, except for basophils. This is to be expected as this is a small population. The correlations between Leukoflow and microscopical differentiation are also quite good although generally lower compared to the LH750.
Table 5. Comparison of the LH750 cell counter, microscopic differentiation, and Leukoflow for the 5-part diff on the normal samples (n = 100)
Cell counter (x)/flow (y)
Microscope (x)/flow (y)
Microscope (x)/Cell counter(y)
y = ax+b
y = ax+b
y = ax+b
Correlation coefficient (r) and slope and offset of the orthogonal regression line (y = ax + b) are given.
0.988x − 1.702
0.880x + 4.862
0.891x + 6.630
1.010x + 0.885
0.863x + 4.123
0.855x + 3.222
1.014x + 0.210
0.812x + 2.839
0.796x + 2.623
1.016x + 0.020
0.855x + 0.294
0.836x + 0.284
1.815x − 0.079
0.466x + 0.506
0.237x + 0.335
Reproducibility of Cellular Classification by Automatic Cell Counter, Microscope, and Leukoflow
The reproducibility of leukocyte differentiation by the automatic cell counter and by Leukoflow was compared by analyzing a normal blood sample 10 times by both methods, performed on three separate days, on different blood samples (Table 6). Except for the absolute WBC, where the coefficient of variation (CV) of Leukoflow is larger than that of the LH750, and the lymphocytes, where both methods have similar CVs, Leukoflow has the lowest CV for all other cell classes. The LH750 is more precise for determining absolute WBC whereas Leukoflow is more precise in differentiating neutrophils, monocytes, eosinophils and basophils. All CVs of the leukocyte populations identified by Leukoflow are <5% except for neutrophilic precursors (Neu preI and Neu Imm), basophils, plasma cells, and blasts. These are small populations in normal blood, which probably explains the increased variation. Microscopical myeloid precursor counts suffer even more from increased variation (20, 21).
Table 6. Reproducibility of cellular classification according to the cell counter (LH750) and Leukoflow, respectively
A normal blood sample was measured 10 times with both techniques on three separate days, performed by different technicians on different blood samples. The coefficient of variation (CV%) is given for the absolute leukocyte count and leukocyte differentiation.
Neu pre I
CD4 pos T-cells
CD4 neg T-cells
Comparison of Automatic Cell Counter, Microscopy, and Leukoflow on Abnormal Samples
Next, we compared the results from 102 abnormal samples (Fig. 2 and Table 7), demonstrating that Leukoflow performs well for the five-part differentiation in samples with aberrant white blood cell numbers. Leukoflow discriminates two categories of granulocyte progenitors (Neu imm and Neu PreI, Plot 10), based on CD16 and CD71 expression, thus providing an estimate of myeloid left-shift. The amount of immature granulocytes identified by Leukoflow is higher than with microscopic differentiation (combining bands, metamyelocytes, myelocytes, pro-myelocytes), indicated by a slope of 2.68 for the regression line (Table 7). When the LH750 indicated presence of normoblasts, the second staining with CD45 and DRAQ5 was performed to identify and quantify NRBC. The correlation between microscope and Leukoflow for the NRBC counts was good (r = 0.96; y = 0.950x+0.041, n = 32). Samples with abnormal leukocyte populations related to different diseases are illustrated in Figures 3A–3T. These examples illustrate that Leukoflow can identify and classify abnormal cell populations or trigger additional investigations in aberrant samples.
Table 7. Correlation when comparing the results of Leukoflow and microscopic differentiation on the abnormal samples (n = 102)
Microscope (x)/flow (y)
y = ax+b
Correlation coefficient (r) and slope and offset of the orthogonal regression line (y = ax+b) are given.
1.00x − 1.84
1.02x − 2.43
1.23x + 0.12
0.99x − 0.02
0.90x + 0.27
2.68x − 2.50
0.60x + 0.17
0.95x + 0.04
Comparison of Blast Cells: Leukoflow Versus Microscope
Reliable blast counting by microscopy is a known problem due to significant inter-observer variability. We identified blasts immunologically in our Leukoflow protocol as CD45 low, CD34 pos cells (Fig. 3B). All samples that were scored with ≥1% blasts either by microscopy or Leukoflow (n = 15) were compared for their blast counts (Table 8). In five samples no blasts are detected by Leukoflow, indicated in gray. In these five discrepant cases, we noted striking differences between the two observers that performed the microscopic differentiation. Additional reviewing of clinical information of these patients revealed underlying disease unlikely to present with blasts, such as stable chronic lymphocytic leukemia, an infection with Epstein-Barr virus or stable chronic myelomonocytic leukemia (CMML). These clinical data suggest no blasts are present in these samples and are in accordance with the results of the Leukoflow analysis. Three samples were also analyzed by routine flowcytometry (Table 8), demonstrating blast counts concordant with Leukoflow.
Table 8. Blast percentage in all samples with ≥1% blasts scored by two independent microscopic differentiations, the average of these, by Leukoflow and by normal routine flow cytometry (when performed; n.d. = not determined), and clinical diagnosis
Gray-shaded cases have lower blast counts by Leukoflow compared with microscopy.
Flow cytometry is a powerful analysis technique, well established in leukemia and lymphoma diagnostics. When performed in a robust, fixed setting using a validated antibody cocktail and protocol this technique can also be applied as a routine test for dWBC. We developed a flow cytometric leukocyte differentiation protocol suitable for replacing part of the microscopic dWBCs performed in a routine laboratory with 24 h/7 days service. Our aim was to create an extended flow cytometric dWBC for samples flagged for review by the hematology analyzer. We studied published protocols and aimed to maximize the clinically relevant information obtained from a “five colors single tube” approach. We aimed to include the 5-part differentiation, all lymphocyte subsets, plasma cells, a CD34-based blast count, a CD14-based monocyte count and an immature myeloid cell count based on absence of CD16, with addition of CD71 as proliferation marker. An additional flow cytometric staining was developed for normoblast detection. We felt this combination would find most clinical acceptance for replacing the microscopic dWBC, adding new possibilities not available with microscopic dWBC. In general our approach was successful and Leukoflow performs very well on normal blood samples. The reference values obtained on our set of normal adult blood samples are shown in Table 9. We are aware that a “five colors single tube” has limitations in the number of identifiable populations and this can be a disadvantage in particular with pathological samples. If aberrant populations are present, when gating or discriminating of populations is difficult or when basophil counts are increased, critical review of the Leukoflow results is crucial and additional investigations such as microscopy or conventional flow cytometry should be considered.
Table 9. Description of values for all absolute cell counts (109/L) measured in adult normal blood samples (n = 92) by Leukoflow
aCD3+/CD8+ T-lymphocytes are approximated in Leukoflow by taking CD3+ CD4− T-cells.
Plasma cells abs
Neutro Pre I abs
Neutro Immature abs
Neutro Mature abs
T cells abs
CD4 T cells abs
CD8 T cells absa
B cells abs
NK cells abs
The use of CD138 allowed us to detect plasma cells in blood. We chose not to use CD38 as this marker can also be expressed on T-, B-, and NK cells (22). We combined CD138 with CD45 in one fluorescence channel and titrated them to stain plasma cells more intense than other leukocytes. In titrating experiments in bone marrow, containing a significant number of plasma cells, this was indeed the case (Fig. 3N). Only two blood samples contained over 1% plasma cells, so we did not compare plasma cell counts of Leukoflow and microscopy. Figure 3M illustrates a plasma cell leukemia in which the plasma cells were not separated from the lymphocytes and basophils, preventing accurate determination of the amount of plasma cells. Plasma cells generally do not express any of the used lymphocyte markers, with sometimes the exception of CD19. If CD138 is not highly expressed they will appear in our protocol as basophils. In these cases additional flow cytometry is needed for proper quantification of plasma cells. More samples with plasma cells need to be included in future studies to confirm this.
Reliable microscopical classification of blast cells is a known problem, even for experienced technicians, illustrated by discrepant blasts counts between the two microscopical observers (Table 8). For more accurate blast definition we included CD34. We are aware that a significant proportion of blasts, especially in AML, is CD34 negative. In Leukoflow CD34 negative blasts may appear as basophils when they are negative for CD3 or CD19. In Figures 3R–3T we have included an additional sample diagnosed with a precursor B-ALL, illustrating the CD34−, CD19+ blasts are visible as CD45 weak in quadrant B1 of Plot 3. More experience with a larger dataset is needed to further establish the accuracy and potential pitfalls of blast enumeration by Leukoflow.
Identification of immature granulocytes is important when reviewing blood samples from patients with potential malignancies, significant infections, or inflammatory disease. Identification of immature cells based solely on morphology, is difficult and subject to significant variation, rendering the clinical utility of the classical “band count” very limited (20, 21). With Leukoflow, granulocyte maturation is quantified by the amount of CD16 and CD71 expression. For immature granulocytes, similar correlations and slopes as published elsewhere were found when comparing microscopic with immunological “left-shift” by Leukoflow (Leukoflow: r = 0.76, slope 2.7; Bjornsson et al.: r = 0.61, slope 2.4; Cherian et al.: r = 0.75/0.81, slopes 1.68/1.57) (12, 14). All immunological methods, whether based on absence of CD16 solely or combined with CD33/64 expression, defining immature myeloid cells, give systematically higher numbers compared to microscopy. We checked the clinical context of all samples where the immature cell count was >5% higher by Leukoflow (n = 30) and found that the majority (n = 23, 76%) of these patients had clinical conditions explaining the presence of myeloid progenitors, such as leukemia, pneumosepsis, or chemotherapy. The remaining seven samples were two cases with B-CLL, three oncology patients (lung- and breast cancer), one patient with arthalgia, and one patient whose clinical information was unavailable. The numbers for left-shift by flow cytometry are different from the traditionally reported number of bands and pro-, meta-, and myelocytes, regardless of which flow cytometric protocol is used. Therefore, we think future studies should focus on the use of this parameter in clinical practice, resulting in clinical validation of immunological left shift as a robust alternative for the traditional band count. In several samples with significant amounts of immature neutrophils the distinction between eosinophils and immature myeloid cells is difficult (Fig. 3G), due to the fact that we did not include a positive marker for eosinophils. However, we believe that proper detection of the immature myeloid cells is the most important clinical diagnostic requirement for these cases, which is fulfilled by Leukoflow.
Three other antibody cocktails for a complete flow cytometric dWBC are reported, each with different composition and gating strategy (12–14). Correlations between the dWBC performed by microscopy, cell counter and Leukoflow (Tables 5 and 7) are comparable to those in literature, with the poorest correlation for basophils (12–14). For erythroblasts the correlations reported here and by Bjornsson are both excellent (12).
Bjornsson et al. differentiated all nucleated cells in 11 categories using six markers and DRAQ5 staining with a 5-color flow cytometer (12). Important differences with Leukoflow are that their cocktail includes CD203 to detect basophils, but cannot discriminate between T and B-cells, and uses CD36 to identify monocytes and platelets. The main advantages of Leukoflow over this cocktail are the more complete characterization of lymphocyte subsets and the inclusion of CD34 to aid blast detection. We targeted our strategy at leukocyte identification and did not consider including platelets.
Faucher and colleagues discriminate 12 different cell populations using a 6-marker/5-colors protocol (13). Significant differences are their use of CD294 to positively identify basophils, and their use of CD2 to detect blasts and mature cells of the T-lineage. Use of CD2 can be a distinct advantage to detect T-ALL with CD34- CD3- blasts. Unfortunately we did not encounter a T-ALL in our cohort. As discussed by Faucher, their description of the lymphocyte subsets is poor. NK and T-cells cannot be discriminated and are not enumerated, in contrast to Leukoflow. In addition they do not use CD34 to aid blast detection. NRBCs and plasma cells are not detected either. Interestingly, they present data on detection of inflammatory syndromes, such as acute bacterial infection, by determining CD16 staining on monocytes. By including a SS versus CD34/16/56 plot, pre-selected on monocytes this is also possible using the Leukoflow cocktail.
The most extended antibody cocktail reported for a single tube so far is that of Cherian et al., who describe an 8-color/10-antibody cocktail including Hoechst staining to detect NRBCs (14). Strong points of their approach are the inclusion of CD34/117 for more robust blast detection, resulting in good correlations with microscopy. However, there is still a significant group of leukemias negative for these markers. Furthermore, CD33/64 is used for positive definition of monocytes and eosinophils, and CD123 for basophils. Unfortunately, there is no positive marker for T-cells included. As many clinical laboratories involved in first-line routine diagnostics are currently operating without, or with a 4/5/6-color flow cytometer, the added complexity and financial aspect of an 8-color flow cytometer may be a disadvantage compared to a 5-color cocktail.
Compared with other published flow cytometric dWBC methods, the Leukoflow assay contains the most markers to identify lymphocyte subsets. Using CD3, CD19, CD16, CD56, and CD4 we provide a complete lymphocyte subset typing, except for the double positive CD3 and CD8 cells. These are approximated by the CD3 positive and CD4 negative cell population which in most cases equals CD3 and CD8 positive cells. The results for all lymphocyte subsets obtained with Leukoflow in normal adult samples compares well with published reference values (23) (Table 9). We are currently performing clinical validation studies to determine the additional diagnostic value of the accurate enumeration of all these populations in well defined patient groups.
Recently, the first flow cytometry routine application where flow cytometric leukocyte differentiation was successfully integrated in the workflow of a hematology laboratory was published (24). Samples flagged by the hematology analyzer are analyzed by flow cytometry before microscopy. This approach reduces the number of microscopical differentiations, manual hands-on time, and turn around time. The 5-color flow cytometric cocktail used in their approach is from Faucher and colleagues (13). It would be highly interesting to test this using the other known flow cytometric dWBC methods. It is crucial to compare these methods on the same extended sample-set, containing a wide range of abnormal samples. This highly interesting aspect is subject of ongoing studies
The absolute WBC counts from the LH750 hematology analyzer and Leukoflow analysis correlated very well and the variation in WBC measured by the LH750 is the lowest (Table 6). In our dataset this allows the use of the absolute WBC of the hematology analyzer and to perform the Leukoflow analysis without flowcount beads. Only in cases where >5% of normoblasts are reported by the hematology analyzer, use of flowcount beads is necessary to obtain absolute leukocyte counts corrected for normoblasts. Uses of tubes with pre-aliquotted counting beads or constant volume acquisition are alternatives to ensure quantitative accuracy of absolute cell counts (25). Each laboratory should evaluate whether a single or dual-platform approach works best in their setting.
The rate of introduction of new flow cytometers with more colors has increased the last years. For flow cytometers with more than five colors, there are several options to extend the Leukoflow method. The antibody cocktail reported here can be used on a 10-color flow cytometer (1 antibody per channel), which requires a less complex gating strategy. For a 6-color instrument we recommend to separate CD34 from CD16 and CD56 and move it to the additional channel, combined with CD117. This way blast detection can be more robust. In addition, discrimination between CD34 low blasts and CD16 low myeloid precursors can possibly be improved. Alternatively, Leukoflow could be complemented with myeloid markers such as CD13 and CD33 combined in the extra channel. This aids the lineage identification of blasts. Another alternative could be allocation of CD138 in a separate channel instead of combining with CD45. This can improve detection and quantification of plasma cells. Also, the inclusion of CD8 or HLA-DR for better description of the lymphocyte populations can be of added value. All of these options require adaptations in the measurement protocol and gating strategy that need to be tested, but they can be performed as an extension of the existing Leukoflow gating strategy.
An alternative technique to facilitate manual dWBC is automated microscopy: an automated microscope with a digital camera and software for cell recognition (26). Advantages over manual microscopy include increased speed and computer-aided cell classification. Identification of cells is morphology-based and the amount of cell populations identified is the same as with manual dWBC. In addition, the counting statistics of automated microscopy (100–400 cells) remain low compared to flow cytometry (26–28). Leukoflow is more robust due to the higher cell numbers and can identify significant more cell populations using immunological cell definitions, both distinct advantages of flow cytometry over microscopy for dWBC.
In addition to accurate identification of cell populations, Leukoflow has to be easy to perform and interpret so that diagnostic information required for adequate treatment of patients can be reported rapidly and accurately. Leukoflow enables easy identification of the different populations as most cell populations automatically fall into their gates. Positioning of gates requires minor adjustments which can possibly be automated in the future. In collaboration with clinicians, we aim to determine the most suitable form to report Leukoflow results after interpretation by the laboratory specialists Clinical Chemistry. Due to small sample volume and antibody titration, reagent costs are less than three euro per sample. The washing step that we performed after lysis is not required to reduce background staining but only to avoid prolonged exposure of stained cells to the ammonium chloride lysis buffer, and can be omitted when samples are measured direct after staining. Alternatively Leukoflow also works well with Versalyse lysing solution (Beckman Coulter, #A09777), which does not require washing either. We tested the Leukoflow protocol on several bone marrow samples, which demonstrated the assay is also suitable for use with bone marrow samples (Fig. 3N and data on file). In collaboration with clinicians in our hospital, we plan studies in different patient groups to clinically validate the parameters determined by Leukoflow and to determine where these data provide additional diagnostic information for use in clinical practice. Taken together we think flow cytometric dWBC as performed by Leukoflow is a highly interesting and promising technique for routine clinical laboratories.