Peripheral blood MDS score: A new flow cytometric tool for the diagnosis of myelodysplastic syndromes

Abstract

Background

Myelodysplastic syndromes (MDS) are a heterogeneous group of hematopoietic disorders diagnosed using morphologic and clinical findings supported by cytogenetics. Because abnormalities may be subtle, diagnosis using these approaches can be challenging. Flow cytometric (FCM) approaches have been described; however the value of bone marrow immunophenotyping in MDS remains unclear due to the variability in detected abnormalities. We sought to refine the FCM approach by using peripheral blood (PB) to create a clinically useful tool for the diagnosis of MDS.

Methods

PB from 15 patients with MDS was analyzed by multiparametric flow cytometry using an extensive panel of monoclonal antibodies. Patterns of neutrophil antigen expression were compared with those of normal controls (n = 16) to establish light scatter and/or immunophenotypic abnormalities that correlated with MDS. A scoring algorithm was developed and validated prospectively on a blinded patient set.

Results

PB neutrophils from patients with MDS had lower side scatter and higher expression of CD66 and CD11a than did controls. Some MDS PB neutrophils demonstrated abnormal CD116 and CD10 expression. Because none of these abnormalities proved consistently diagnostic, we sought to increase the power of the assay by devising a scoring system to allow the association of multiple abnormalities and account for phenotypic variations. The PB MDS score differentiated patients with MDS from controls (P < 0.0001) in the test set. In a prospective validation, the PB MDS score successfully identified patients with MDS (sensitivity 73%, specificity 90%).

Conclusions

FCM analysis of side scatter and only four additional immunophenotypic parameters of PB neutrophils using the PB MDS score proved more sensitive than standard laboratory approaches and may provide an additional, more reliable diagnostic tool in the identification of MDS. © 2005 Wiley-Liss, Inc.

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal hematologic disorders characterized by ineffective hematopoiesis and eventual progression (1). The diagnosis of MDS requires a multidisciplinary approach involving hematologic, morphologic, and cytogenetic analyses. However, the diagnosis may be difficult to render because more than 50% of patients present with one or fewer cytopenias and only approximately 40% of patients demonstrate cytogenetic abnormalities (2). Several recent studies have suggested a role for flow cytometric (FCM) analysis in the evaluation of MDS (3–16). Most of these studies have focused on complex immunophenotypic evaluation of the bone marrow. Stetler-Stevenson et al. (4) systematically investigated the potential diagnostic utility of flow cytometry of bone marrow in MDS patients and convincingly demonstrated that, by using an extensive antibody panel and examining various cell types, patients with MDS can be identified. Further, they observed that the evaluation of larger numbers of myeloid FCM parameters allowed additional sensitivity for identifying MDS without compromising specificity. These findings were verified in more recent studies conducted by Wells et al. (3) who confirmed the utility of multiparameter flow cytometry of bone marrow in the diagnosis of MDS. In addition, this group demonstrated that a thorough evaluation of the complex FCM changes present in the bone marrow of patients with MDS has prognostic significance. These and other previous studies argue for the diagnostic and, potentially, prognostic utility of flow cytometry in the diagnosis of MDS.

In the present investigation, we performed FCM analysis of peripheral blood (PB) neutrophils (PMNs). Specifically, we assess multiple phenotypic parameters in an objective fashion to test the hypothesis that multiparameter FCM analysis of PBNs may provide a useful adjunct in the diagnosis of MDS. Further, we describe and validate a scoring system that can easily be implemented in the clinical FCM laboratory as an additional tool to facilitate the diagnosis of MDS.

MATERIALS AND METHODS

Patients

Pretreatment PB samples from 15 patients with a diagnosis of MDS (n = 13) or of a myeloproliferative/myelodysplastic disorder (n = 2), as defined by World Health Organization (WHO) criteria (17), were collected with informed consent (Table 1). All samples were reviewed and classified by the hematopathology service at the Hospital of the University of Pennsylvania. The control group consisted of 16 samples selected from excess material submitted to the hematologic laboratory for routine complete blood cell counts from patients without MDS. For the validation study, blinded samples were provided by one of the authors (A.B.) from de-identified samples (n = 31) of patients with cytopenias and/or hematologic dysplasia noted on the PB smear. After un-blinding of diagnoses, 11 patients in the validation set were classified as having MDS by a board-certified hematopathologist according to WHO diagnostic criteria.

Table 1. Patient Characteristics (Test Set)

Sample	Diagnosisa	Age (years)	Cytogenetics	IPSS	MDS score
a World Health Organization classification. CMML, chronic myelomonocytic leukemia; MDS, myelodysplastic syndrome; MDS-U, MDS unclassified (case 1, MDS with fibrosis); NA, not available; RA, refractory anemia; RAEB, refractory anemia with excess blasts; RARS, refractory anemia with ringed sideroblasts; RCMD, refractory cytopenia with multilineage dysplasia. b 43–45,X,del(X)(p22),del(1)(p32p36.3),add(3)(?q21),−5,add(5)(p13),−7,−10, del(11)(q23),del(12)(p13)−13,−19, −20,−22,+2−4mar[cp25]. c Two of 65 cells with del(7)(q22q32). d Calculation of International Prognostic Scoring System requires cytogenetics.
1	MDS-U	78	Normal	0.5	7
2	RCMD	71	Normal	0.5	5
3	RA	50	Normal	0	4
4	RAEB-2	59	Complexb	3	8
5	RARS	58	Normal	0	4
6	RCMD	61	Normal	0.5	3
7	RCMD	79	Normal	0.5	2
8	RA	79	Normal	0.5	7
9	RAEB-2	73	Normalc	2	0
10	RAEB-1	77	Normal	1	3
11	CMML-1	83	NA	NAd	4
12	CMML-2	66	Normal	1.5	6
13	RCMD	63	Normal	0	3
14	RCMD	47	i(17)(q10)	1	4
15	RCMD	75	del (5)(q15q35)	1.5	4
	Mean	67.93
	Standard deviation	11.20
	Control (n = 16)
	Mean	54.13
	Standard deviation	18.01

Sample Preparation

All samples were collected in tubes that contained ethylenediaminetetraacetic acid and processed within 24 h of collection. PB samples were subjected to whole blood red blood cell lysis with 1× ammonium chloride (NH₄CL concentration of 0.15 mol/ml) lysing solution for 10 min at room temperature. After lysis, cells were washed twice with phosphate buffered saline and resuspended in staining buffer (phosphate buffered saline with 0.5% bovine serum albumin). The PB leukocytes were then stained according to manufacturers' recommendations with a panel of antibodies, directly conjugated to fluorochromes, against antigens expressed at various stages of normal and abnormal myeloid differentiation (18) or against antigens that have been reported to be altered in MDS (3–16) (Table 2). All antibodies were obtained from BD Biosciences (San Jose, CA, USA) with the exception of anti-CD64 phycoerythrin (PE; Caltag Laboratories, Burlingame, CA, USA) and anti-CD116-PE (Immunotech, Marseille, France). All antibodies used are routinely quality controlled and evaluated for appropriate concentration by the clinical FCM laboratory of the Hospital of the University of Pennsylvania. Concentrations used have been found to be greater than or equal to saturating levels. After staining, samples were washed two times with staining buffer, resuspended in staining buffer, and analyzed immediately.

Table 2. Peripheral Blood MDS Antibody Panel*

	Fluorochrome
	FITC	PE	PERCP	APC
* Antigens were selected based on previous observations as cited (references in parentheses). APC, allophycocyanin; FITC, fluorescein isothiocyanate, MDS, myelodysplastic syndrome; PE, phycoerythrin; PerCP, peridinin chlorophyll protein.
Tube 1			CD45
Tube 2	CD71 (7)	CD10 (4,9)	CD45	CD32 (15)
Tube 3	CD16 (3,4,7)	CD116 (5,11,20)	CD45	CD13 (3,4,7)
Tube 4	CD18 (14,15)	CD11a (15)	CD45	CD11b (3,4,15)
Tube 5	CD59 (13)	CD55 (13)	CD45	CD56 (3,4)
Tube 6	CD90 (12)	CD117 (6)	CD45	CD34 (3)
Tube 7	CD43 (14)	CD64 (4)	CD45	HLADR (3,7)
Tube 8	CD66 (14,16)	CD44 (8,14,16)	CD45	CD33 (3,7)

Flow Cytometric Analysis

Data were acquired on a FACSCalibur (BD Biosciences) that was calibrated daily using eight peak SPHERO Rainbow Calibration Particles (Spherotech, Inc., Libertyville, IL, USA) that monitored variables including fluorescent intensity and peak width to ensure optimal and standardized instrument performance. Compensation was adjusted using cells stained with CD8 and fluorescein isothiocyanate (FITC), CD8-PE, CD8 peridinin chlorophyll protein (PerCP), and CD8 allophycocyanin (APC) (all directly conjugated and processed in an identical fashion to the test samples). Data were analyzed with Flow Jo Software (Treestar, San Carlos, CA, USA). A four-color panel was used, with FITC, PE, and PerCP excited by the 488-nm argon laser and detected in FL1, FL2, and FL3, respectively, and allophycocyanin excited by the 635-nm red diode laser and detected in FL4. In each preparation, a minimum of 10,000 total events was collected. CD45-PerCP was included in each tube to allow identification of PMNs using CD45 versus side scatter (SSC) gating (19) (Fig. 1). Geometric mean fluorescent intensity (G-MFI) was derived for the autofluorescence control and for each antibody. The G-MFI for each antibody was then corrected for the autofluorescence [(test G-MFI − auto-G-MFI)/(auto-G-MFI)] to generate the corrected G-MFI [adapted from Maynadie et al. (7)]. To assess PMN SSC we draw from the experience of Wells et al. (3) who demonstrated that myeloid cell granularity can be represented as the difference between lymphocytic and myeloid granularity. We quantitatively expressed the corrected PMN SSC as [(mean PMN SSC)/(mean lymphocytic SSC)].

Figure 1. Gating strategy. Lymphocyte and granulocyte gates were drawn based on CD45 versus SSC characteristics. For the autofluorescence control and each antibody, the G-MFI was derived. The corrected G-MFI was calculated as illustrated. Mean SSC was calculated for the lymphocyte and granulocyte gates and the corrected granulocyte SSC was calculated.

Statistical Analysis

Mann-Whitney U test was used to assess statistical differences between groups. All statistical calculations were performed with Analyse-it for Microsoft Excel (Leeds, UK; http://www.analyse-it.com/).

RESULTS

Alterations in SSC and Immunophenotypic Patterns in PBNs From MDS Patients

In preliminary studies, the extensive panel listed in Table 2 was evaluated on PB monocytes, blasts (identified by CD45/SSC gating) and PMNs; however, changes in PMNs were most informative and thus were the focus of this study. PMNs have been described to be immunophenotypically, morphologically, and functionally abnormal in patients with MDS (15, 16). By using an extensive panel of immunophenotypic markers, chosen as outlined above, we identified several significant, quantifiable alterations in PMNs from MDS patients. As has previously been reported in bone marrow (3, 4), PB granulocytes demonstrated a significant decrease in SSC compared with controls, reflecting hypogranularity (corrected SSC of 5.81 ± 1.73 vs. 7.94 ± 1.01, P = 0.001; Fig. 2). In addition, PMNs from MDS patients had significantly higher expression of CD66 (corrected CD66 G-MFI 42.35 ± 24.75 vs. 18.78 ± 5.56, P = 0.0003) and CD11a (corrected CD11a G-MFI 98.66 ± 33.02 vs. 69.39 ± 19.03, P = 0.008; Figs. 3 and 4) than did those from controls. Variable abnormalities of CD10 (previously demonstrated in bone marrow granulocytes) (9) and CD116 (as reported in previous studies) (5, 11, 20) were noted in some samples. Specifically, one patient demonstrated marked loss of CD10, one patient demonstrated CD116 expression greater than two standard deviations (S.D.) above the mean of normals, and two patients demonstrated CD116 expression less than 2 S.D. below the mean of normals (data not shown).

Figure 2. SSC abnormalities in MDS. A: CD45 versus SSC plot of representative MDS and control samples. B: MDS PMNs have a significantly lower SSC than those of normal controls (P = 0.001).

Figure 3. CD66 expression of PMNs from MDS patients versus controls. A: CD66 and autofluorescence are demonstrated for a representative patient with MDS and for a control sample. B: CD66 expression was significantly higher for MDS samples than for controls (P = 0.0003).

Figure 4. CD11a expression of PMNs from MDS patients versus controls. A: CD11a and autofluorescence are demonstrated for a representative patient with MDS and for a control sample. B: CD11a expression was significantly higher for MDS samples than for controls (P = 0.008).

Development of the PB MDS Score

Because MDS are a heterogeneous group of diseases, it is unlikely that any parameter taken in isolation would allow discrimination of all patients with MDS. Thus, a scoring system was devised as a method to associate the heterogeneous immunophenotypic data into an easier- to-interpret format. Based on our observation of phenotypic changes in the PMNs of MDS patients, this scoring system evaluates only PMNs. To create the score, statistically significant differences (SSC, CD11a expression, CD66 expression) were assigned 1 point for variation from 1 to 2 S.D. from the mean of normal and 2 points for variations greater than 2 S.D. from the mean of normal. In addition, points were assigned for other described abnormalities such as loss of CD10 or abnormal CD116 expression. Because abnormalities in CD10 and CD116 were rare in the test set, 2 points were given for loss of CD10 or abnormal CD116 expression (2 S.D. above or below the mean of normal). The sum of points for each sample was defined as the PB MDS score. The PB MDS score was significantly higher in patients with MDS than in controls (4.27 ± 2.09 vs. 0.56 ± 0.81, P < 0.0001; Fig. 5).

Figure 5. Development of the PB MDS score. The PB MDS score was calculated by combining the noted abnormalities identified in patients with MDS as described. As illustrated, the PB MDS score was significantly higher in patients with MDS than in normal controls (P < 0.0001).

Validation of the PB MDS Score

To validate the accuracy of the PB MDS score, the diagnostic approach was applied to a blinded series of 31 patients with cytopenias and/or hematopoietic dysplasia. After completion of the scoring, diagnoses were unblinded to reveal that 11 patients in the validation set had a confirmed diagnosis of MDS according to WHO criteria and the remaining 20 patients did not (Table 3). Scores from patients with MDS were significantly higher than those for controls (4.36 ± 1.80 vs. 2.05 ± 1.23 for patients without MDS, P = 0.0005; Fig. 6A). A receiver operating curve analysis performed on the validation set (Fig. 6B) illustrated that a cutoff higher than 3 maximizes the sensitivity and specificity of the PB MDS score. Using a cutoff point higher than 3 to define MDS, three patients with MDS were misclassified and two controls were misclassified. Using this cutoff, in the validation set, the MDS score performed with a sensitivity of 73% and a specificity of 90%.

Table 3. Validation of the Peripheral Blood MDS Score

Sample no.	Diagnosis	MDS score
a Each of these patients had at least one cytopenia, but none had documented MDS. AML, acute myelogenous leukemia; MDS, myelodysplastic syndrome; MDS-U, MDS unclassified (case 1, MDS with fibrosis); RA, refractory anemia; RAEB, refractory anemia with excess blasts; RCMD, refractory cytopenia with multilineage dysplasia.
MDS patients
1	RAEB-1	1
2	MDS with fibrosis	3
3	RCMD	3
4	RAEB-2	4
5	RCMD with myeloproliferative features	4
6	MDS-U	4
7	RA	5
8	RCMD	5
9	RCMD	5
10	MDS-U, recurrent posttransplantation	6
11	RCMD	8
	Mean	4.36
Control patientsa
1	AML	0
2	Iron deficiency anemia	0
3	Anemia of chronic disease	0
4	Anemia of chronic disease	1
5	Anemia of unknown etiology	1
6	Anemia of unknown etiology	1
7	Anemia of chronic disease	2
8	Anemia of unknown etiology	2
9	Anemia of chronic disease	2
10	Chronic idiopathic myelofibrosis	2
11	AML	2
12	Anemia of chronic disease	2
13	Anemia of chronic disease	3
14	Anemia of chronic disease	3
15	Anemia associated with chemotherapy for carcinoma	3
16	Thrombocytopenia of unknown etiology	3
17	Chronic idiopathic myelofibrosis	3
18	Anemia of chronic disease	3
19	Anemia of chronic disease	4
20	Chronic lymphocytic leukemia	4
	Mean	2.05

Figure 6. Validation of the PB MDS score. The MDS score was validated in a set of unknown samples taken from patients with cytopenias and/or dysplasia on PB smear. A: In the validation set, patients with MDS have a higher PB MDS score than do those without MDS (P = 0.0005). B: Receiver operating curve (ROC) analysis demonstrates that using a cutoff higher than 3 to identify patients with MDS minimizes false-positive and false-negative results in the validation set.

DISCUSSION

Although the role of flow cytometry is well established for the diagnosis of other disorders of myeloid lineage such as acute myeloid leukemia, the utility of this approach for MDS has not been well defined. Recently, several studies have attempted to clarify the role of several individual markers and of various analysis techniques, applied primarily in the bone marrow, in the diagnosis of MDS. In this report, we have described a unique approach that uses SSC and immunophenotypic data derived from PMNs to create a scoring algorithm that can accurately distinguish patients with MDS from controls. In this study, we have shown that multiparameter FCM analysis of PB using the PB MDS score may be a useful adjunct in the diagnosis of MDS. Further, we have demonstrated that the evaluation of only five PMN FCM parameters may be at least as sensitive as existing objective (cytogenetic) testing and perhaps even more specific for the identification of patients with MDS.

Because flow cytometry allows the rapid detection of numerous phenotypic alterations, it is a well-suited modality to analyze a disease process as heterogeneous as MDS. Previous investigators including Stetler-Stevenson et al. (4) and Wells et al. (3) demonstrated that the ability to integrate multiple pieces of immunophenotypic data allows for more reliable identification of patients with MDS by flow cytometry. Stetler-Stevenson et al. (4) evaluated multiple cells types (blast, myeloid, monocytic, erythroid, megakaryocytic) with an extensive panel of (>10) antibodies. Using this method, they highlighted the contributions of flow cytometry to the conventional tools used to diagnose MDS such as morphology and cytogenetics. Wells et al. (3) used a similar approach that evaluated more than 10 phenotypic markers of blast, myeloid, and monocytic cells to derive an FCM scoring system that can be used to not only diagnose MDS but also provide prognostic information. The method of analysis in both studies was qualitative and complex, requiring interpretation by an experienced reader familiar with the multitude of changes in various cell types that can be present in MDS. Mayndie et al. (7) addressed some issues of complexity in interpretation by providing a numeric representation of antigen expression (fluorescence ratio of the marker of interest and the autofluorescence on the cell type of interest). They demonstrated that quantitative expression of several antigens including CD34, CD36, and CD71 on various cell types is informative because it correlates with prognostically significant variables in MDS such the International Prognostic Scoring System score and the French-American-British category.

The benefit of our approach in comparison with those of previous investigators is that it maintains the ability to integrate the heterogeneity of phenotypic changes observed in MDS and maintains a level of simplicity that would allow it to be easily integrated into a clinical FCM laboratory. In the present study, we focused on PB PMNs and combined data about SSC and only four informative PMN immunophenotypic variables to create the PB MDS score, a multivariate predictor. Not only is this assay simpler than previous approaches, it also is relatively noninvasive compared with previous studies because it is performed on PB. Using the PB MDS score we were able to separate patients with MDS from non-MDS controls with a high level of accuracy (P ≤ 0.005). We then validated this scoring system in a test set of patients with cytopenias and/or hematopoietic dysplasia, characteristics associated with MDS, to determine the potential of our system to differentiate MDS from other hematologic conditions.

A thorough understanding of the biologic and clinical significance of the changes noted in PMNs of patients with MDS requires further inquiry and is not the subject of this study; however, we can make some speculations given what is currently understood about the biology of MDS. The decreased SSC noted in patients with MDS has been described previously in bone marrow studies (3, 4) and likely represents the morphologically appreciated hypogranularity of PMNs in patients with MDS. CD66 is a member of the carcinoembryonic antigen family of proteins that is expressed on various tissue types. On neutrophils, CD66 is correlated with activation and with increased CD11/CD18-mediated cellular adhesion (21). Increased CD66 and CD11a expression may therefore be related to increased PMN activation in patients with MDS. This finding correlates with previous studies suggesting that MDS PMNs have an activated phenotype (14). Patients with early MDS appear to have accelerated apoptosis of myeloid cells, whereas more advanced categories of disease have a decrease in apoptotic activity and an increase in cell proliferation (22). The consistent finding of alterations in CD10 expression in MDS in this and other studies (9) may reflect abnormal apoptosis because CD10 may be a marker of apoptosis (23, 24). CD116, the α-chain of the granulocyte-macrophage colony-stimulating factor receptor, has been shown to decrease in MDS PMNs (5, 11, 20). In the present study, we found two patients with low CD116 expression and one patient with high CD116 expression. Alterations in this critical and complex cytokine pathway may have important implications in the development of MDS (1, 25, 26). Although specimen processing could have resulted in some of the observed changes, especially in the expression of activation markers, it is clear that these changes preferentially occur in the PMNs of MDS patients. The previously reported changes in activation antigen expression by PMNs from MDS patients might reflect a lower threshold for activation, thus making this population more susceptible to activation induced by processing. As with many biomarker studies, the reason for the phenotypic change may not be entirely clear, but this and other observations might be exploited as an aid in the diagnosis of MDS.

In summary, we have described a PB scoring system based on FCM analysis of PMNs that may provide a useful adjunct to existing testing to more accurately diagnose MDS. The PB MDS score was derived from a noninvasive, relatively simple immunophenotypic approach that performed well in a validation set including patients with some clinical characteristics suspicious for MDS. With newly available technology that allow simultaneous evaluation of six or more immunophenotypic parameters, this assay, in its current form, can be easily converted to a two-tube analysis to be performed in the clinical FCM laboratory. Further, multicenter studies are needed to confirm these preliminary data and to assess correlations between the PB MDS score and clinical parameters relevant in the evaluation of patients with MDS. When combined with other clinical and laboratory parameters, the sensitivity may be increased, resulting in a new paradigm for MDS diagnosis.

Acknowledgements

Jonni Moore is supported in part by grant P30 CA16520 from the National Institutes of Health, and Adam Bagg is supported by a grant from the Leukemia and Lymphoma Society of America (7000-02).