Combined flow cytometric assessment of CD45, HLA-DR, CD34, and CD117 expression is a useful approach for reliable quantification of blast cells in myelodysplastic syndromes

Authors

  • Alex F. Sandes,

    1. Division of Hematology, Department of Clinical and Experimental Oncology, Escola Paulista de Medicina-Universidade Federal de São Paulo (UNIFESP-EPM), São Paulo, Brazil
    Search for more papers by this author
  • Daniella M. B. Kerbauy,

    1. Division of Hematology, Department of Clinical and Experimental Oncology, Escola Paulista de Medicina-Universidade Federal de São Paulo (UNIFESP-EPM), São Paulo, Brazil
    Search for more papers by this author
  • Sergio Matarraz,

    1. Centro de Investigación del Cáncer (IBMCC, CSIC-USAL) and IBSAL, Servicio de Citometría and Departamento de Medicina, Universidad de Salamanca, Salamanca, Spain
    Search for more papers by this author
  • Maria de Lourdes L. F. Chauffaille,

    1. Division of Hematology, Department of Clinical and Experimental Oncology, Escola Paulista de Medicina-Universidade Federal de São Paulo (UNIFESP-EPM), São Paulo, Brazil
    Search for more papers by this author
  • Antonio López,

    1. Centro de Investigación del Cáncer (IBMCC, CSIC-USAL) and IBSAL, Servicio de Citometría and Departamento de Medicina, Universidad de Salamanca, Salamanca, Spain
    Search for more papers by this author
  • Alberto Orfao,

    1. Centro de Investigación del Cáncer (IBMCC, CSIC-USAL) and IBSAL, Servicio de Citometría and Departamento de Medicina, Universidad de Salamanca, Salamanca, Spain
    Search for more papers by this author
  • Mihoko Yamamoto

    Corresponding author
    1. Division of Hematology, Department of Clinical and Experimental Oncology, Escola Paulista de Medicina-Universidade Federal de São Paulo (UNIFESP-EPM), São Paulo, Brazil
    • Disciplina de Hematologia e Hemoterapia, Departamento de Oncologia Clínica e Experimental, Universidade Federal de São Paulo – UNIFESP. R. Botucatu 740, V. Clementino, 04023-900, São Paulo, SP, Brazil
    Search for more papers by this author

  • How to cite this article: Sandes AF, Kerbauy DMB, Matarraz S, MLLF Chauffaille M, López A, Orfao A, Yamamoto M. Combined flow cytometric assessment of CD45, HLA-DR, CD34, and CD117 expression is a useful approach for reliable quantification of blast cells in myelodysplastic syndromes. Cytometry Part B 2013; 84B: 157–166.

Abstract

Background.

Quantification of bone marrow (BM) blasts by cytomorphology is essential for the diagnosis of myelodysplastic syndromes (MDS). Owing to its subjectivity and the potential impact of dysplastic features on accurate identification of blast cells, more objective approaches are required, multiparameter flow cytometry (MFC) being a particularly promising approach in this regard. However, no consensus exists about the optimal combination of markers and strategy to be used.

Methods.

BM blast counts from 74 MDS patients were evaluated by morphology versus four different MFC phenotypic criteria: “CD34+”, “CD34+ and/or CD117+”, “CD34+, and/or CD117+HLA-DR+”, and “CD34+ and CD117+HLA-DR+ plus CD64+CD14−/lo” cells. For each criterium, the percentage of blasts was calculated using either all BM nucleated cells or non-erythroid CD45+ cells as denominator.

Results.

The number of “CD34+ and/or CD117+HLA-DR+”cells showed the highest correlation and agreement with morphological counts, only a minor proportion of cases being misclassified by MFC vs. morphology for the >5% and >10% classification thresholds. In turn, a CD34+ phenotype was insufficient to correctly identify and quantify blasts. Conversely, usage of non-erythroid BM cells as denominator, or inclusion of “CD34+ and/or CD117+HLA-DR+ plus CD64+CD14−lo” cells were both associated with overestimated blast counts.

Conclusions.

Quantification of “CD34+ and/or CD117+HLA-DR+” cells (from all nucleated BM cells) by MFC is an efficient method for the enumeration of blasts in MDS. However, caution should be taken with replacing morphology by MFC blast counts; its combined use may rather provide complementary information increasing the accuracy and reproducibility of BM blast cell counts in these patients. © 2013 International Clinical Cytometry Society

Current criteria for the diagnosis, classification, and prognostic evaluation of myelodysplastic syndromes (MDS) are largely based on conventional morphological and cytogenetic analyses (1). Among other variables, accurate identification and enumeration of blast cells remain essential both for the differential diagnosis with acute leukemia and the prognostic classification of MDS, increased numbers of blast cells reflecting a higher risk of disease progression, leukemic transformation, and death (2).

However, cytomorphological enumeration of blast cells frequently remains subjective as it mainly depends on the multiple variables such as the expertise of the observer, the precise criteria used to define myeloblast cell counts and the quality of bone marrow (BM) smears (3–5). Most importantly, dysplastic features may also hamper correct identification of blast cells in MDS leading to understimated blast cell counts (4, 5). For years, the search for additional criteria that could contribute to a more objective identification and quantification of BM blast cells in MDS has been pursued, but it still remains a challenge nowadays. Among other criteria, quantification of CD34+ cells by multiparameter flow cytometry (MFC) has been proposed and evaluated, but it proved to underestimate morphological blast cell counts in a significant number of patients (6–11). More recently, usage of additional markers expressed by immature myeloid cells (e.g., CD117) in combination with CD34, has also been proposed for the identification of blast cells among either BM nucleated cells or white cells (8, 9, 12), in an attempt to increase the precision of blast cell counts by MFC. Nevertheless, although the combination of redundant markers to assess BM blast cell counts is reasonable and well accepted this concept mostly derives from working conferences and expert opinions. Furthermore, the impact of immature monocytic precursors on BM blast cell counts has never been appropriately assessed, as these cells lose expression of CD34 and CD117 early during hematopoietic differentiation. Despite this, to the best of our knowledge no study has been reported so far in which the different immunophenotypic strategies proposed for the identification and quantification of BM blast cells have been simultaneously assessed and compared with conventional morphological blast cell counts.

In the present study, we investigated the correlation between blast cell counts obtained by eight different MFC strategies versus conventional cytomorphology, in a series of 74 MDS patients evaluated in parallel at two different centers.

METHODS

Patients

A total of 74 patients (36 males and 38 females; mean age of 72 ± 10 years, ranging from 43 to 90 years) diagnosed with MDS according to the World Health Organization (WHO) 2008 criteria (13) at the Universidade Federal de São Paulo (UNIFESP, São Paulo, Brazil; n = 42) and the University Hospital of Salamanca (Salamanca, Spain; n = 32), were studied. In parallel, EDTA-anticoagulated normal and reactive BM samples (n = 10) from an age-matched group of subjects were collected. All patients gave their informed consent before entering the study, in line with the guidelines of the local Ethics Committees and the Helsinki Declaration, after the study was approved by the local Institution Review Boards. According to the WHO criteria, patients were distributed as follows: refractory cytopenia with unilineage dysplasia, 16 cases—14 refractory anemias (RA) and 2 refractory thrombocytopenias (RT); refractory anemia with ringed sideroblasts (RARS), 10 patients; refractory cytopenia with multilineage dysplasia (RCMD), 22; refractory anemia with excess of blasts type 1 (RAEB-1), 11; RAEB-2, 13 cases; isolated del(5q) and unclassifiable MDS, one patient each; at the moment of the study, one RAEB-2 patient showed transformation into acute leukemia with 22% BM blasts.

Cytomorphological Studies

BM samples were obtained by first-pull aspiration. Part of the sample was used to prepare BM smears and another part was collected into tubes containing EDTA as anticoagulant for further MFC assays. BM smears were stained with May-Grünwald-Giemsa and cell morphology was analyzed by conventional light microscopy (Nikon Eclipse E/200 microscope, Nikon, Tokyo, Japan) using a 100x/1.25 oil immersion objective. The acceptable quality of samples was defined by the presence of spicules, megakaryocytes, and other hematopoietic precursors in BM slides, according to the guidelines of the International Council for Standardization in Hematology (14). All BM smears were considered as being appropriate for morphological analysis; no case of hypocellular MDS or MDS with fibrosis was included in the study. In order to evaluate the reproducibility of morphological blast cell counts, samples were evaluated by two independent experienced cytomorphologists and both ≥ 500 nucleated cells and ≥30 megakaryocytes were analyzed in each BM smear. Both erythroid and granulocytic dysplasia were defined by the presence of ≥10% BM cells of the respective lineage with morphological alterations; presence of ≥15% ringed sideroblasts was also considered as a diagnostic criterium for erythroid dysplasia. Morphological features used for the definition of myeloblasts were those proposed by the International Working Group on Morphology of MDS (15), and the blast cell percentage was determined using the overall number of BM nucleated cells as denominator.

Immunophenotypic Identification and Enumeration of Blast Cells

Whole BM sample aliquots (2 × 106 cells in 50–100 μL per test) were stained with saturating amounts of monoclonal antibodies (MAb) directed against cell surface markers using a stain-lyse-and-then-wash direct immunofluorescence technique, as previously described in detail (16). Briefly, samples were incubated for 15 min in the dark at room temperature (RT). The following MAb were systematically used in two distinct 4-color combinations: (1) HLA-DR–fluorescein isothiocyanate (FITC); clone L243; Becton/Dickinson Bioscences (BDB), San Jose, CA, CD117–phycoerythrin (PE; clone 104D2; BDB), CD45–peridinin chlorophyll protein (PerCP-Cy5.5; clone 2D1; BDB), and CD34–allophycocyanin (APC; clone 8G12; BDB), and; (2) CD36–FITC (clone FA6.152; Immunotech, Marseille, France), CD64–PE (clone 22, Immunotech), CD45–PerCP-Cy5.5, and both CD34-APC plus CD14-APC (clone MØP9, BDB). After staining, 2 ml of FACS lysing solution (BDB) diluted 1/10 (vol/vol) in distilled water was added to each tube and after gentle mixing, samples were incubated for 10 min in the dark at RT, in order to lyse the non-nucleated red cells. Then, cells were centrifuged (5 min at 540g) and the cell pellet washed in 4 ml of PBS. Finally, cells were resuspended in 0.5 ml of PBS and immediately run in a FACSCalibur flow cytometer (BDB).

Data acquisition was performed immediately after sample preparation was completed using two FACSCalibur flow cytometers (one at each site) and the CellQUEST software program (BDB). For each sample, information was acquired about a total of >5 × 104 nucleated cells/tube. For data analysis, the Paint-A-Gate Pro (BDB) and the Infinicyt (Cytognos, SL, Salamanca, Spain) software programs were used.

Immunophenotypic identification of BM blast cells was performed according to four different criteria; for each criterium, the percentage of blast cells was calculated using both the total number of nucleated BM cells (T) and the number of non-erythroid (CD45+) nucleated cells (N), as denominator. In detail, the criteria used for the definition of blast cells (T1 and N1 to T4 and N4, respectively) were as described below. For the T1 and N1 strategies, blast cells were defined as those CD34hi and CD45lo/int cells showing low-to-intermediate forward (FSClo/int) and sideward (SSClo/int) light scatter characteristics, following the ISHAGE guidelines (Fig. 1A) (17). In turn, the T2 and N2 approaches, considered blast cells to correspond to those cellular events showing a “CD34hi and/or CD117+” and CD45lo/int phenotype, independently of their FSC and SSC features (Fig. 1B). The T3 and N3 methods identified “CD34hi and/or CD117+HLA-DR+” cells showing CD45lo/int expression as corresponding to blast cells (Fig. 1C). Finally, the T4 and N4 strategies defined blast cells as those cellular events corresponding to the immature monocytic precursors (CD64hi/CD14neg/lo/CD34neg) in addition to those cells defined as blast cells with the T3 and N3 criteria (Fig. 1D). Whenever detected, CD34+CD117++ mast cells and CD34+/CD117/SSClo B-cell precursors (hematogones), were systematically excluded from the blast cell population identified on phenotypic grounds. All percent values were calculated after excluding cell debries and cell doublets from the analysis.

Figure 1.

Representative bivariate dot plots illustrating the different MFC gating strategies used for the identification and quantification of bone marrow blast cells in patients with myelodysplastic syndromes and normal/reactive BM. Panels in column A illustrate the usage of CD34 alone (T1 strategy) for the identification of blast cells (red dots); in column B, blasts are identified following the T2 strategy (CD34+ and/or CD117+ BM cells; red dots); column C depicts the usage of the T3 strategy (CD34+ and/or CD117+HLA-DR+ BM cells; red dots) and; in column D the T4 strategy is examplified, with inclusion of CD64++CD14neg/loCD34 monocytic precursors in the blast cell gate in addition to CD34+ and/or CD117+HLA-DR+ BM cells identified by the T3 strategy.

Nucleated erythroid cells were identified on the basis of their unique light scatter characteristics (low side and forward scatter) and negative or low expression of CD45. An adittional tube containing CD45 (FITC; clone 2D1; BDB), CD34 (PE; clone 8G12; BDB), and the DRAQ5 DNA dye (Biostatus, Leicestershire, UK) was performed in 25 MDS cases to evaluate whether the number of erythroblasts was correctly assessed and detached from large platelets and unlysed red blood cells by the scatter/CD45 gating strategy used in the absence of a DNA dye. The difference between the number of erythoblasts with and without DRAQ5 was irrelevant and the number of nucleated erythroid cells obtained with both strategies was identical (r2 = 0.99) (Fig. 2).

Figure 2.

Correlation of multiparameter flow cytometry enumeration of nucleated erythroid precursors based on the DRAQ5 DNA dye vs. only their scatter/CD45 profile in a subset of 25 MDS patients.

In order to evaluate the reproducibility of MFC blast cell counts, all MDS cases were assessed by two independent expert observers, each one from a distinct center.

Statistical Methods

For all parameters evaluated, mean values and their standard deviation (SD) as well as median and range values were calculated. To determine the degree of correlation and agreement between blast cell percentages as assessed by the distinct immunophenotypic criteria versus cytomorphology, the Pearson correlation (r2) and the Bland-Altman tests were used, respectively. The limits of agreement of the Bland-Altman test were set as the mean difference between morphologic and immunophenotypic blast counts ± 2SD. Sensitivity was calculated as TP/(TP + FN) (TP: true positive cases; FN: false negative cases), specificity as TN/(TN + FP) (TN: true negative cases; FP: false positive cases), and efficiency as (TP + TN)/(total cases); positive and negative predictive values were calculated as TP/(TP + FP) and TN/(TN +FN), respectively. Receiver operator curves (ROC) were used to establish the most efficient cut-off values for each immunophenotypic strategy to predict cytomorphological blast cell counts. P values < 0.05 were considered to be associated with statistical significance. All statistical analyses were performed using the SPSS 13.0 (SPSS, Chicago, IL) and the Medcalc (Mariakerke, Belgium) software programs.

RESULTS

The overall mean cytomorphological blast cell count assessed in normal/reactive BM samples (n = 10) was of 1.52% (range: 1.0%–2.1%) (Table 1). When comparing the distinct immunophenotypic criteria, the “T3” (CD34+ and/or CD117+HLA-DR+) strategy showed the highest degree of correlation and agreement with morphological blast cell counts (r2 = 0.89; mean difference of 0.2%), whereas blast counts were underestimated by the T1 and N1 strategies (mean difference of −0.6% and −0.4%, respectively) and overestimated once the T2, N2, T4, N3, and N4 criteria were applied (mean overestimated levels of 1.3%, 1.8%, 1.2%, 0.6%, and 1.7%, respectively) (Table 1). Usage of CD45+ cells (non-erythroid BM cells) vs. all nucleated BM cells as denominator, also led to overestimation of blast cell counts with all phenotypic approaches evaluated. Noteworthy, very similar results (P > 0.05) were found when the Sao Paulo and the Salamanca cohorts of patients were separately analyzed (data not shown).

Table 1. Normal/Reactive and MDS BM: Quantification of BM Blast Cell Percentages by Eight Different Multicolor Flow Cytometry (MFC) Strategies (T1, T2, T3, T4 and N1, N2, N3, N4) Versus Conventional Morphology
 Normal/reactive BM (n = 10)MDS patients (n = 74)
   Degree of agreement with morphology  Degree of agreement
Method of analysis% of blasts*Correlation with morphology (r2)Mean difference ± 1SDLimit of agreement n (%)% of blasts*Correlation with morphology (r2)Mean difference ± 1SDLimit of agreement n (%)
  1. Results expressed as mean values ± one standard deviation and range between brackets. MFC, multiparameter flow cytometry; SD, standard deviation. Blast percentages were calculated using either the total number of nucleated BM cells (T) or the number of non-erythroid (CD45+) nucleated (N) cells as denominator. Blasts were defined as: CD34hi/CD45dim/int/FSClo/int/SSClo/int cells (T1 and N1), CD34hi and/or CD117+ and CD45lo/int (T2 and N2), CD34hi and/or CD117+/HLA-DR+/CD45lo/int (T3 and N3) or as CD64hi/CD14neg/lo monocytic precursors plus those blasts defined by the T3 and N3 criteria.

Morphology1.52 ± 0.4% (1–2.1%)   4.5 ± 4.8% (0.1–22%)   
MFC        
T10.9 ± 0.2% (0.6–1.3%)0.84−0.6% ± 0.2%10 (100%)2.7 ± 2.8% (0.3–15%)0.68−1.9% ± 3.6%68 (92%)
T22.8 ± 0.7% (1.4–3.6%)0.591.3% ± 0.5%10 (100%)6.1 ± 4.4% (1–22.3%)0.721.5% ± 3.4%67 (90%)
T31.7 ± 0.4% (1–2.1%)0.890.2% ± 0.2%10 (100%)4.7 ± 3.8% (0.8–16.3%)0.770.2% ± 3%69 (93%)
T42.8 ± 0.6% (1.6–4%)0.731.2% ± 0.4%9 (90%)5.5 ± 4.2% (1–19.5%)0.761.0% ± 3.2%69 (93%)
N11.1 ± 0.3% (0.7–1.6%)0.63−0.4% ± 0.3%10 (100%)3.6 ± 3.9% (0.3–19.5%)0.67−0.9% ± 3.6%67 (90%)
N23.3 ± 1% (1.6–4.6%)0.371.8% ± 0.9%10 (100%)7.7 ± 5.5% (1–26.5%)0.723.3% ± 3.9%69 (93%)
N32.1 ± 0.5% (1.1–2.9%)0.610.6% ± 0.4%10 (100%)5.9 ± 4.7% (0.9–20.9%)0.751.5% ± 3.4%69 (93%)
N43.3 ± 0.8% (1.8–4.3%)0.681.7% ± 0.6%10 (100%)7.1 ± 5.1% (1–21.2%)0.752.5% ± 3.6%70 (94.5%)

The overall mean cytomorphological blast cell count observed among the 74 MDS BM samples analyzed was of 4.5% (range: 0.1%–22%), 24 cases presenting >5% BM blast cell counts by morphology. In turn, the mean blast cell percentage obtained by MFC—using all nucleated BM cells as denominator (“T” criteria)—varied according to the specific criteria used: 2.7% (range: 0.3%–15%), 6.1% (range: 1%–22%), 4.7% (range: 0.8%–16%), and 5.5% (range: 1%–19.5%) for the T1 (CD34+), T2 (CD34+ and/or CD117+), T3 (CD34+ and/or CD117+HLA-DR+), and T4 (CD34+ and/or CD117+HLA-DR+ plus CD64hiCD14−/lo) criteria, respectively. In turn, the blast cell count by MFC using non-erythroid CD45+ BM cells as denominator, was significantly higher (P < 0.05) for all four MFC methods: 3.6% (range 0.3%–19.5%), 7.7% (1%–26%), 5.9% (0.9%–21%), and 7.1% (1%–21%) for the N1, N2, N3, and N4 strategies, respectively (Table 1). The mean blast cell percentages obtained with the T3 (CD34+ and/or CD117+HLA-DR+) criteria were those being closest (P > 0.05) to the morphological counts: 4.6% vs. 4.5%, respectively.

As could be expected, “T3” (CD34+ and/or CD117+HLADR+) also corresponded to the MFC strategy showing the strongest correlation and degree of agreement with morphology: r2 = 0.77 with a mean overall difference of 0.2% vs. r2 ≤ 0.76 (T1, r2 = 0.68; T2, r2 = 0.72; T4, r2 = 0.76; N1, r2 = 0.67; N2, r2 = 0.72; N3, r2 = 0.75 and; N4, r2 = 0.75) with mean overall differences of −0.9% to 3.3% (mean difference of −1.9%, −0.9%, 1.5%, 3.3%, 1.5%, 1.0%, and 2.5% for the T1, N1, T2, N2, N3, T4, and N4 strategies, respectively) (Table 1). From the other methods, T1 and N1 (both based on CD34+ cell counts only) systematically underestimated the morphological counts (mean difference of −1.9% and −0.9%, respectively), whereas the T2, N2, T4, and N4 criteria were associated with higher blast cell counts than those observed on cytomorphological grounds, with a mean overestimated value of 1.5%, 3.3%, 1.0%, and 2.5%, respectively. Similarly, the N3 method also overestimated the blast cell percentages obtained by morphology (mean difference of 1.5%) (Table 1 and Fig. 3).

Figure 3.

Degree of agreement between bone marrow (BM) blast cell counts by flow cytometry (MFC) and conventional morphology in MDS (n = 74). Horizontal lines represent mean of differences (−) and ± 2SD (---) of MFC blast cell percentage differences from morphology. Panels A, C, E, and G represent MFC counts obtained when all nucleated BM cells were used as denominator with strategies T1, T2, T3, and T4, respectively; in turn, panels B, D, F and H represent MFC blast cell percentages when only the non-erythroid CD45+ BM cells were considered as denominator with the N1, N2, N3, and N4 strategies, respectively. The percentages inside the Bland-Altman plots indicate the number of cases inside the 95% limits of agreement.

Comparison of morphological blast cell counts obtained by two independent expert observers showed a strong correlation (r2 = 0.72) and degree of agreement (mean overall difference = 1.1%) between them. Nevertheless, MFC blast cell counts also obtained by two independent observers showed a higher correlation than morphology, for the eight immunophenotypic strategies investigated (T1, r2 = 0.97; T2, r2 = 0.98; T3, r2 = 0.96; T4, r2 = 0.96; N1, r2 = 0.97; N2, r2 = 0.97; N3, r2 = 0.93; N4, r2 = 0.94) and degree of agreement (mean overall difference of T1 = 0.3%; T2 = −0.001%; T3 = 0.7%; T4 = 0.15%; N1 = 0.3%; N2 = −0.1%; N3 = 1%; N4 = 0.3%).

In a second step, we evaluated the utility of MFC vs. cytomorphology for the quantification of blast cells at levels that are critical for the classification of the disease (e.g., 5%, 10%, and 20% blasts used as cut-off values for the diagnosis of RAEB-1, RAEB-2 and transformation to acute leukemia, respectively). Once again, MFC definition of blast cells as “CD34+ and/or CD117+HLA-DR+″ cells was associated with the greatest efficiency: 85% and 92% for the diagnosis of RAEB-1 and RAEB-2, respectively (most informative cut-off values of 4.2% and 8% CD34+ and/or CD117+/HLA-DR+ cells, respectively) (Table 2 and Fig. 4). From the other MFC criteria, those based exclusively on CD34 counts (T1 and N1 strategies) presented the lowest cut-off values for the identification of both RAEB-1 and RAEB-2 cases (2% and 2.8% blasts with a lower overall efficiency of 76% and 78% for the T1 criteria, and 4.9% and 5.1% blasts with an overall efficiency of 82% and 88% for the N1 method, respectively). Those MFC strategies, which used non-erythroid CD45+ cells as denominator, were systematically associated with higher cut-off values vs. those found for those strategies using the total number of BM cells as denominator. However, regarding the specific identification of RAEB-2 cases the T2, T3, and T4 strategies showed very similar results.

Figure 4.

Receiver operating curve (ROC) analysis of the efficiency of the different MFC strategies used to calculate the percentage of BM blast cells in MDS. In panels A and C, overall nucleated BM cells were used as denominator (strategies T1, T2, T3, and T4), whereas in panels B and D non-erythroid BM cells (N1, N2, N3, and N4 strategies) were used as denominator to calculate blast cell percentages when morphologic blast cell counts are >5% and >10%, respectively.

Table 2. Performance of Eight Different Flow Cytometry Strategies for the Quantification of BM blast cells versus Conventional Morphology in the Classification of MDS Patients (n = 74) into Distinct WHO Subtypes
Method of analysisAUCCut-off valueSensitivitySpecificityPPVNPVEfficiency
  1. AUC, area under the ROC curve; PPV, positive predictive value; NPV, negative predictive value. Blast percentages were calculated using either the total number of nucleated BM cells (T) or the number of non-erythroid (CD45+) nucleated (N) cells as denominator. Blasts were defined as: CD34hi/CD45lo/int/FSClolo/int/SSClo/int cells (T1 and N1), CD34hi and/or CD117+ and CD45lo/int (T2 and N2), CD34hi and/or CD117+/HLA-DR+/CD45lo/int (T3 and N3) or as CD64hi/CD14neg/lo monocytic precursors plus those blasts defined by the T3 and N3 criteria.

Morphologic blast cell count > 5% (RAEB-1)
T10.812%79%76%59%88%76%
T20.825.1%79%74%58%88%74.3%
T30.874.2%83%86%74%92%85%
T40.875.1%78%86%72%89%81%
N10.794.9%58%98%82%82%82%
N20.787.7%67%86%70%84%79.7%
N30.855.6%79%86%70%89%82.4%
N40.846%78%82%62%88%75.6%
Morphologic blast cell count > 10% (RAEB-2)
T10.812.8%77%82%44%94%78%
T20.898.4%85%93%69%96%90%
T30.908%77%95%72%95%92%
T40.898.1%83%95%72%96%92%
N10.805.1%69%93%64%93%88%
N20.889.6%85%90%61%96%88%
N30.886.4%84%88%58%96%86%
N40.879.9%83%92%63%96%89%

Only 1/44 MDS cases presented >20% blast cells by morphology (22%); the blast cell count in this case was of 15%, 18%, 18%, and 18% for the T1, T2, T3, and T4 strategies, respectively, and of 20%, 21%, 21%, and 21% for the N1, N2, N3, and N4 approaches. In turn, false positive MDS cases showing >20% blasts by phenotype but not by morphology were found with the T2 (n = 1 case), N2 (n = 3 cases), and N4 (n = 2 cases) criteria, but never with the T3 approach.

DISCUSSION

Quantification of the number of blast cells in the BM of MDS patients is a key parameter for the classification of the disease, as well as for the differential diagnosis between MDS and acute leukemia(13). Currently, the gold-standard to determine blast cell counts is conventional cytomorphology. However, this is a relatively subjective technique that frequently lacks precision, particularly in the low range counts typically found among low-grade MDS, due to the fact that ≤500 cells are evaluated from which ≤100 would systematically correspond to blast cells (18). Previous studies have proven that in such cases flow cytometry-based enumeration of CD34+ cells could represent a reproducible approach for the quantification of normal and neoplastic hematopoietic progenitor and precursor cells (8, 19, 20). However, MFC enumeration of CD34+ BM cells is not universally accepted for quantification of blast cells since it systematically provides underestimated blast counts (13, 21). Consequently, other alternative immature hematopoietic cell phenotypes have been proposed (e.g., CD34+HLA-DR+CD11b) (8, 22). Despite this, to the best of our knowledge no study has been reported so far in which the number of BM cells with such multiple immature phenotypes are simultaneously enumerated and results compared with those derived from morphological blast counts.

In this study, we investigated the correlation between four different phenotypic criteria used for MFC identification of blast cells in BM samples from MDS patients versus morphological blast cell counts; in addition, for each of the four approaches, percent values from both all nucleated cells and only CD45+ BM cells, were calculated. Overall, our results confirm that usage of CD34+CD45lo alone is not sensitive enough to identify the blast cell compartment in MDS BM, as it systematically underestimated morphological blast counts, independently of the denominator used. This is most probably due to the fact that CD34 is a hematopoietic precursor cell antigen whose expression decreases during differentiation before the loss of the morphological “blast” appearance; consequently, a small proportion of myeloblasts are already CD34-negative (23). This is also true for both AML and MDS blasts, which frequently show loss of CD34 expression in the context of an unequivocal blast cell morphology (7, 24).

Conversely, when we considered the “CD34+ and/or CD117+” immunophenotype, MFC was associated with an overestimation of blast cell numbers compared with morphology. This could be due to inclusion of relatively mature CD117+ myeloid precursors, which have already lost morphological features of a blast cell; in fact, CD117 expression is typically retained over the normal and altered myeloblast stage, in cells differentiating to different myeloid lineages, such as the neutrophil (e.g., promyelocytes) and erythroid (e.g., basophilic erythroblasts) lineages (8, 23, 25). Thus, usage of a marker that would restrict the phenotype of CD117+CD34 cells to a more immature stage, could potentially contribute to a better phenotypic definition of the blast cell morphological appearance. Actually, the immunophenotypic criteria for MFC identification of blast cells that more closely correlated with the morphological counts was that based on a “CD34+ and/or CD117+HLA-DR+” profile, using the total number of BM nucleated cells as denominator (the so-called T3 criteria), with a negligible mean difference between both approaches. Within the “CD34+ and/or CD117+HLA-DR+” phenotype, CD34+ myeloblasts are included together with CD34CD117+HLA-DR+ neutrophil- (CD45int/SSClo/int) and erythroid-committed precursors (CD45loSSCvery-lo), as well as early CD117loHLA-DRhi monocytic and dendritic cell precursors (25–27). Additional inclusion of CD64++CD14−/loCD34 monocytic precursors in the blast cell counts did not improve the correlation between MFC and morphological blast cell counts, but it led to overestimation of MFC versus morphological blast counts, in a significant fraction of MDS cases (data not shown). Overall, these results are in line with general consensus recommendations from the literature, which highlight the need to use more than one monoclonal antibody associated to precursor cells (CD34, CD117 or HLA-DR) to properly identify blasts by MFC (1, 9); at the same time they further prove the increased reproducibility of the “CD34+ and/or CD117+/HLA-DR+” criteria. Nevertheless, caution should be taken when using MFC blast cell counts in an individual basis, as a replacement for the morphological counts, reinforcing the concept that MFC should be interpreted as an independent tool to assess blast cell counts. Moreover, parallel sorting experiments are required to definitively demonstrate the morphologic blast cell nature of CD34+ and/or CD117+/HLA-DR+ BM cells in MDS and its absence in their normal/reactive counterpart. The CD34+HLA-DR+CD11b phenotype for blast cell quantification was not evaluated in this study, as it employs only one immature marker (e.g., HLA-DR) within the CD34+ cell compartment, leading to underestimation of CD117+HLA-DR+CD34 immature cells.

Several studies have indicated that blast cell percentages based on non-erythroid CD45+ cells versus all nucleated BM cells may be more consistent because of the potential variable impact of red cell lysing reagents (e.g., ammonium chloride) in the number of nucleated erythroid precursors due to uncontrolled lysis of red cells (9). Although usage of the total number of nucleated cells as denominator for MFC blast cell enumeration is proposed in the current version of the European LeukemiaNet (ELN) recommendations for immunophenotyping of MDS, this working group recognizes the need for a prospective study to comparatively evaluate the usage of “non-erythroid CD45+ cells” versus “all nucleated BM cells,” to define the denominator of choice for the enumeration of blast cells by MFC (12). Here, we show that the usage of non-erythroid CD45+ cells as denominator is frequently associated with overestimated blast cell percentages and a lower correlation with morphological blast counts, versus percent values calculated from the total nucleated BM cells; therefore, total nucleated cells appear to be a more robust denominator (vs. nucleated CD45+ cells), unless the conventional morphological criteria is also changed to reflect blast cell percentages from white nucleated BM cells, which seems unlikely.

Some groups have proposed the use of nuclear dyes to eliminate residual mature non-nucleated and large platelets for a correct quantification of cell populations in peripheral blood (PB) and BM, using total nucleated cells as denominator (28, 29). Parallel comparison of scatter versus DRAQ5 gating was performed in only one third of the MDS cases here analyzed. However, the strong correlation observed for the gating strategies based on removal of cell debries and cell doublets by light scatter parameters and the DRAQ5-DNA dye gating suggests that the use of nuclear dyes is not strictly required to properly enumerate blast cells by flow cytometry.

Another relevant clinical question relates to the efficiency of MFC counts for the classification of MDS into RAEB-1, RAEB-2, and acute leukemia. Although previous studies have described a lower CD34+ threshold in MDS cases with high numbers of blasts by morphology (7, 27, 30, 31), thus far no cut-off value for disease classification has been proposed and adopted to quantify blast cells in MDS by MFC. As could be expected, the “CD34+ and/or CD117+HLA-DR+” criterium to define percent blast cells from all BM nucleated cells by MFC, also showed the higher efficiency for the identification of both RAEB-1 and RAEB-2 cases. However, it should be noted that the most informative cut-off values were slightly lower than those currently used by cytomorphology: 4.2% versus 5% and 8% versus 10%, respectively. Such slight discrepancy could be due to the fact that compared with cytomorphology MFC may be associated with a slight underestimation of blast cell counts in samples where these are increased in the BM particles. This is because such BM particles are negatively selected during the flow cytometric measurements, particularly when no disaggregation procedures are used, and potential contamination of the sample with PB exists. In fact, evaluation of PB contamination of BM samples still remains a concern. Despite several different approaches and mathematical formulas have been proposed to evaluate PB contamination of BM samples (22, 32, 33), most of them cannot be easily implemented in routine diagnostics or they only indirectly and semiquantitatively evaluate the potential contamination with PB, based on the size of the mature cell compartments (e.g., mature CD16++ or CD13++CD11b++ neutrophils) and other BM-restricted cell compartments (e.g., nucleated red blood cells, mast cells, CD138hi plasma cells). Likewise, for manual morphological analyses, BM samples with PB contamination are not suitable to quantify blast cells by MFC. Further studies are necessary to define the most reliable approach to determine PB contamination of BM samples, which is suitable for routine assessment in diagnostic laboratories.

In summary, here, we show that quantification of blast cells by MFC based on a “CD34+ and/or CD117+HLA-DR+” phenotypic profile is a robust tool for the classification of MDS, such criterion significantly improving the efficiency of “CD34+” and “CD34+ and/or CD117+” blast cell counts.

Acknowledgements

The authors thank Dr. Tsutomu Oguro, chief of MDS out patient service at HSPE.

Ancillary