307

Address correspondence to Kyungja Han, M.D., Departmentof Clinical Pathology, Catholic University Medical College,St. Mary’s Hospital, Youngdeungpo-gu, Youido-dong 62, Seoul,Korea (South) 150-713; tel 82 2 3779 1297; fax 82 2 7836648; e-mail: hankja@catholic.ac.kr.

Use of an Automated Hematology Analyzer and Flow Cytometryto Assess Bone Marrow Cellularity and Differential Cell Count

Myungshin Kim, Jayoung Kim, Jihyang Lim, Yonggoo Kim, Kyungja Han, and Chang Suk KangDepartment of Clinical Pathology, College of Medicine,The Catholic University of Korea, Seoul, Korea

Abstract. The automated analysis of bone marrow aspirates was performed to estimate the bone marrowcellularity and differential cell count. Total nucleated cell count (TNC) was measured using an automatedhematology analyzer. TNC of the bone marrow correlated well with the bone marrow cellularity estimatedby microscopic examination (r = 0.590, p <0.001). The bone marrow cellularity was readily confirmed as<30%, from 30 to 70%, or >70% according to TNC. Differential count of bone marrow cells was donewith the combination of CD45 monoclonal antibody and propidium iodide using flow cytometry. Excellentcorrelations were obtained for the population distributions of normoblasts, eosinophils, lymphocytes, andblasts between the results of flow cytometry and manual differential counts. The percentage of maturegranulocytes in flow cytometry showed good correlation with the manual percentage of metamyelocytes +neutrophils. The percentage of immature granulocytes by flow cytometry showed good correlation with themanual percentage of blasts + promyelocytes + myelocytes. These results demonstrate that analysis of bonemarrow aspirates using an automated hematology analyzer is valuable in the determination of bone marrowcellularity. Moreover, as flow cytometry provides objective findings for percentages of major cell populations,it can serve as a method to automate bone marrow differentials. (received 8 March 2004; accepted 11 March 2004)

Keywords: bone marrow, differential count, hematology analyzer, flow cytometry, CD45, propidium iodide

Introduction

Microscopic examination of bone marrow has longbeen a major method for diagnostic and therapeuticevaluation of hematologic disease [1]. Whenpathologists observe bone marrow, they judge themarrow cellularity by histologic sections of a biopsyand ascertain the distribution of cells by a manualdifferential count. Though the biopsy is the goldstandard for determining cellularity, determinationof bone marrow cellularity on an aspirate is oftenhelpful because preparing histologic sections takesmore time than an aspirate. Moreover, as thecellularity may differ according to the site, it is betterto examine many sites than only one site.

The manual differential cell count is an essentialpart of this work in order to have an objective record.However, such microscopic examination does notproduce an objective differential count because ofinter-observer differences [2]. The manualdifferential counting method is imprecise due to thesmall number of cells counted, typically not morethan 500 nucleated elements. It is also labor-intensive and time-consuming [3].

Attempts to analyze bone marrow aspiratesusing automated hematology analyzers or flowcytometry have been performed for years, but suchapplications have remained limited since bonemarrow is a heterogenous tissue containing multiplematuration stages of all lineages of hematopoieticcells [4,5]. The availability of fluorochrome-conjugated monoclonal antibodies to variouslineage-specific and lineage-associated cell surfaceantigens provides more objective and quantitativeidentification of the individual cell types found in

0091-7370/04/0300-0307 $1.75. © 2004 by the Association of Clinical Scientists, Inc.

Annals of Clinical & Laboratory Science, vol. 34, no. 3, 2004

308

bone marrow [3,6]. However, these approaches arenot easily adapted to routine use in clinicallaboratories because of the complex analysis softwarethat is required for multidimensional analysis [7].The use of various monoclonal antibodies to identifythe individual cell populations is so expensive thatit has limited routine use. The method is more usefulin routine diagnostic hematopathology when it issimple to perform.

In this study, we estimate the bone marrow cell-ularity using an automated hematology analyzer. Inaddition, we use a flow cytometric method fordifferential counting of bone marrow cells usingCD45 monoclonal antibody and propidium iodide(PI). As a outcome of this study, we are able toprovide objective data including cellularity anddifferential count of bone marrow aspirates prior tomorphologic assessment.

Materials and Methods

Patients. We studied 369 samples of bone marrowfrom patients who had been admitted to theHematology Department of St. Mary’s Hospital. Thesamples (2 ml) of EDTA-anticoagulated bonemarrow were collected and analyzed within 2 hr ofcollection. The patients’ hematological disorderswere acute myeloid leukemia (AML, n = 39), acutelymphoblastic leukemia (ALL, n = 22), chronicmyelogenous leukemia (CML, n = 42), aplasticanemia (AA, n = 43), post-stem cell transplantationstate (BMT, n = 35), incomplete remission state afterchemotherapy (IR, n = 48), and controls (n = 140).The controls consisted of patients in completeremission state after chemotherapy and clinicalspecimens being evaluated for the bone marrowinvolvement of lymphoma. They were restricted tobone marrow aspirates with <5% pathological cellson microscopic examination.

Microscopic examination of bone marrow. Bonemarrow smears were prepared on glass slides, and,after staining with Wright stain, were observed undera microscope. Differential counts by microscopicexamination were used as a reference [8]. Themarrow cellularity was determined by both bonemarrow aspiration and biopsy section. The biopsy

cellularity was used when there was discordancebetween the aspirate and biopsy cellularity.

Analysis by automated hematology analyzer. Weanalyzed the bone marrow aspirates using a Coulter®

GEN·S™ (GEN·S) automated hematology analyzer(Coulter Corp, Miami, FL). EDTA-anticoagulatedbone marrow was diluted 10-fold with normal saline.After the diluted sample was filtered through 53 µm-pore sized nylon mesh, the total nucleated cell count(TNC) was measured.

Flow cytometry for differential count of bone marrowcells. 1. Sample preparation. Differential count ofbone marrow samples using flow cytometry wasperformed as follows: Bone marrow (100 µl) at aconcentration of 4.0-5.0 x 106 cells was mixed with10 µl of CD45, fluorescein isothiocyanate (FITC)(BD Biosciences, San Jose, CA) and incubated for20 min in the dark at room temperature. Two ml ofFACS™ lysing solution (BD Biosciences) was addedand the mixture was incubated for 10 min. Afterwashing twice with normal saline, the sample wasincubated with 10 µl of PI (BD Biosciences) for 10min. The final sample volume was adjusted to 500µl using normal saline.

2. Data acquisition. All flow cytometric measure-ments were carried out within 60 min of samplepreparation. Flow cytometric analysis was performedon a FACSCalibur (Becton-Dickinson Immuno-cytometry System (BDIS), San Jose, CA). Forwardangle light scatter (FSC), side angle light scatter(SSC), and two-color fluorescence signals (green andorange fluorescence at 530 and 585 nm, respectively)were determined for each cell and stored in list modedata files. In toto, 20,000 events were recorded foreach specimen.

3. Data analysis. List mode data files were analyzedusing CellQuest Software (BDIS). As a first step,four groups of cells were gated in the CD45-FITCand PI dot-plotted according to the intensity ofCD45 and PI (R1: CD45 negative, R2: CD45weakly positive & high PI, R3: CD45 weaklypositive & low PI, R4: CD45 strongly positive).Then, each group of cells was dot-plotted again by

Annals of Clinical & Laboratory Science

309

FSC and SSC combination. The cell populationswere further identified by their FSC and SSCintensity. In the cases of AML, ALL, and IR, wedifferentiated the cells into 2 groups, L1 and L2,with L1 defined as CD45 weakly positive or negativecells and L2 as CD45 strong positive cells. Next,each group of cells was dot-plotted by FSC and SSCcombination and identified by the clustering ofpopulations.

Statistical analysis. Numerical values were expressedas mean ± SD; t tests were conducted and Pearson’scorrelation coefficients determined for analysis ofdata. The flow cytometry results and manualmicroscopic results were examined and comparedby regression analysis. The correlation coefficient wasdetermined by SPSS version 9.0. Differences wereconsidered significant at p 0.05.

Results

Cellularity. TNC obtained by the automatedhematology analyzer was 97.4 ± 126.2 x 109/L. TNCof the bone marrow correlated well with the bonemarrow cellularity estimated by microscopicexamination (r = 0.590, p = 0.000). When thecellularity was categorized as <30%, from 30 to 70%,and >70%, there was significant difference of TNCamong the different cellularity groups (33.8 ± 30.0x 109/L vs 83.8 ± 75.8 x 109/L vs 202.8 ± 173.8 x109/L, p <0.001) (Fig. 1).

Clustering of bone marrow populations by CD45-FITC and PI. The nucleated elements wereseparated from debris by CD45-FITC and PI dotplot. The patterns of normal bone marrow werehighly reproducible with 4 separate groups of cells,all easily recognizable by combined CD45-FITC andPI (Fig. 2A). The positions of these cell clusters wereso consistent that the boundaries could be definedusing one set of data and required little or no changefor the analysis of other sets of data even if theywere collected on different days. The cell populationsthat were secondarily gated by FSC and SSCintensity after first gating were identified as follows:R1, normoblasts and immature granulocytes; R2,normoblasts and immature granulocytes; R3,

normoblasts, monocytes, eosinophils, and maturegranulocytes; and R4, lymphocytes, monocytes, andmature granulocytes (Fig. 2B,C). As a result,populations of normoblasts, lymphocytes, mono-cytes, eosinophils, immature granulocytes, andmature granulocytes were classified.

The cell populations of the CML, AA, andBMT cases were well separated using the same gatingsets that were used in controls. On the other hand,in cases of AML, ALL and IR, the 4 groups of cellswere not easily separated by CD45-FITC and PIdot-plot and the subpopulations of each group ofcells could not be correctly identified by the presenceof a lot of blasts. Therefore, we analyzed those casesusing the following modified method: Instead ofgrouping the cells into four categories by CD45-FITC and PI dot plot, we differentiated them intotwo groups, L1 and L2, according to their CD45intensity. The secondarily gated cell populations wereidentified as follows: L1, normoblasts, blasts, andtotal granulocytes; and L2, lymphocytes, blasts, andtotal granulocytes (Fig. 3). As a result, populationsof normoblasts, lymphocytes, total granulocytes, andblasts were classified.

In acute leukemia cases, the cells on scattergramby FSC and SSC were clustered in one population.Typically AML revealed a wider distribution of SSC

Fig. 1. Comparison of total nucleated cell counts (TNC) amongthe different cellularity groups.

Automated analysis of bone marrow aspirates

310

Fig. 2. Flow cytometric analysis of bone marrow aspirate from a patient in the control group after staining with CD45-FITC andPI. In the first step, 4 groups of cells were gated (panel A). In the second step, each group of cells was reanalyzed into separate cellclusters: N, normoblasts; IG, immature granulocytes; MG, mature granulocytes; Mono, monocytes; Eos, eosinophils; and Lym,lymphocytes (panel B). A case with marked eosinophilia shows eosinophils that are easily separated from MG (panel C)

Table 1. Comparison of subpopulation percentages (mean ± SD) obtained by manual differential count and by the flow cytometricmethod in bone marrow aspirates from the total group of 369 cases.

Cell types Manual Flow cytometry correlation pdifferential (%, n = 369) coefficient vs manual

(%, n = 369) (r) differential

normoblasts 19.8 ± 16.9 14.8 ± 12.4 0.775 <0.001lymphocytes 11.4 ± 14.3 10.3 ± 13.8 0.867 <0.001eosinophils 2.0 ± 4.3 2.0 ± 2.8 0.945 <0.001mature granulocytes 42.7 ± 17.0* 44.8 ± 16.0 0.682 <0.001immature granulocytes 11.8 ± 8.5† 19.0 ± 10.8 0.624 <0.001monocytes 3.5 ± 3.8 4.3 ± 2.6 0.463 <0.001blasts 55.8 ± 33.5 57.7 ± 30.1‡ 0.911 <0.001total granulocytes 19.1 ± 25.0§ 21.2 ± 24.3 0.926 <0.001

* metamyelocytes + myelocytes† blasts + promyelocytes + myelocytes‡ counted only in acute myeloid leukemia, acute lymphoblastic leukemia, and incomplete remission cases§ promyelocytes + myelocytes + metamyelocytes + neutrophils

Annals of Clinical & Laboratory Science

08 Kim 307-313 8/10/04, 2:25 PM310

311

Fig. 3. Flow cytometric analysis of two patients with acute leukemia. In the first step, two groups of cells were gated (panel A). Inthe second step, each group of cells was reanalyzed into separate cell clusters: N, normoblasts; Bl, blasts; Lym, lymphocytes; G,total granulocytes. A case of acute lymphoblastic leukemia (panel B). A case of acute myeloid leukemia (panel C). Patient (C)shows a higher SSC level than patient (B).

than ALL did. To differentiate AML from ALL, thegeometric mean value of SSC was useful, beingsignificantly greater in AML than in ALL (198.4 ±77.1 vs 96.7 ± 22.6, p <0.001) (Fig. 4).

Differential counts of bone marrow cells. Table 1presents the correlation data between flow cyto-metric differential data and manual differentialmeasurements. Excellent correlations between the2 methods were obtained in normoblasts, lympho-cytes and eosinophils. Although the correlation formonocytes was low, it was significant (r = 0.463, p<0.001). The percentages of normoblasts differedbetween the 2 methods, with microscopic valuesbeing about 5% higher than the flow cytometryones. The percentage of mature granulocytes in flowcytometry had a better correlation with the manualpercentage of metamyelocytes + neutrophils thanwith the manual percentage of myelocytes +metamyelocytes + neutrophils (r = 0.682 vs r =0.568). The percentage of immature granulocytes

Fig. 4. Comparison of side angle scattergram geometric meanvalues (SSC GEOM) between acute myeloid leukemia (AML)and acute lymphoblastic leukemia (ALL). The bold linerepresents the median value of SSC GEOM, the box representsthe ± SD range of SSC GEOM, and the vertical line represents

the range of individual SSC GEOMs.

Automated analysis of bone marrow aspirates

312

by flow cytometry had the best correlation with themanual percentage of blasts + promyelocytes +myelocytes (r = 0.624, p <0.001). In the cases ofAML, ALL, and IR, the blast count correlated wellbetween the two methods (r = 0.911, p <0.001).The percentage of total granulocytes by flowcytometry correlated well with the sum of promyelo-cytes, myelocytes, metamyelocytes, and neutrophils(r = 0.926, p <0.001).

Discussion

In this study, we analyzed the diluted bone marrowaspiration sample by an automated hematologyanalyzer on the basis that counts of total nucleatedelements using the automated hematology analyzercorrelate significantly with manual counts [2,9].TNC of the bone marrow correlated well with thebone marrow cellularity estimated by microscopicexamination (r = 0.590, p <0.001). By this method,we can easily confirm the bone marrow cellularityas either <30%, from 30 to 70%, or >70% accordingto TNC. The estimated bone marrow cellularity bythe automated hematology analyzer was differentfrom that by microscopic examination in a few cases.Such discrepancy was most prominent in bonemarrow with >70% cellularity, which does not revealany adequate particles on the smeared specimenbecause of bone marrow fibrosis or bone marrowpacked with leukemic cells. The bone marrowaspiration sample itself was unsatisfactory not onlyfor the evaluation of bone marrow cellularity butalso for the diagnosis of bone marrow conditionbecause of the admixing of bone marrow withsinusoidal blood [10]. Except for such cases, theautomated hematology analyzer can be useful for arough estimate of bone marrow cellularity.

In this paper, we evaluated a simple andinexpensive flow cytometric method to classify thelineage and maturity of bone marrow cells. Thismethod does not require specialized handling orprocessing techniques. By combining the CD45expression of a cell with its PI intensity, we effectivelyseparated the basic groups of cells for bone marrowdifferentials. CD45 has different surface expressionsbased on the lineage and maturity of the cell. CD45expression is highest in mature lymphocytes and

monocytes, and erythroid precursors lack the CD45antigen [11]. PI is used to separate the nucleatedelements of bone marrow from red blood cell (RBC)ghosts, debris, lyse-resistant RBC, and platelets [12].We found that the immature granulocytes showedhigher PI intensity than normoblasts and maturegranulocytes. It has been known that the PI intensitywas affected by the permeability, chromatinstructure, and histone content of various cell types[13,14]. We suspect that most of the S- and G2/M-phase cells consist of immature granulocytes and thattheir chromatin pattern is less compact. In the casesof AML, ALL, and IR, the 4 groups of cells werenot easily gated by CD45-FITC and PI dot-plot,and the subpopulations of each group of cells wereincorrectly identified, owing to the presence of manyblasts. Therefore, we analyzed these cases using themodified method described in the results. Theconventional flow cytometric markers of FSC andSSC, which reflect size and granularity, provide thebest visual differentiation for more detailed bonemarrow differential counts [2,3].

We found that results obtained by this techniqueshow strong correlation with the manual bonemarrow differential counts over a wide range ofvalues. Excellent correlations between the twomethods were obtained for the populationdistributions of normoblasts, eosinophils, andlymphocytes. Although it remained significant, thecorrelation for monocytes was low because thepercentage of bone marrow monocytes was low andthere was no case with a high percentage of themonocyte fraction. The percentages of normoblastsdiffered between the two methods, with microscopicvalues being higher than the flow cytometry ones.It is assumed that some of these cells were lost byhemolytic agents. The percentages of maturegranulocytes and immature granulocytes obtainedby flow cytometry gave good correlations with themanual percentages of metamyelocytes + neutrophilsand of blasts + promyelocytes + myelocytes,respectively. In the cases of AML, ALL, and IR, theblast count correlated well between the two methods.The percentage of total granulocytes by flowcytometry correlated well with the sum ofpromyelocytes, myelocytes, metamyelocytes, andneutrophils and was considered to be a normal

Annals of Clinical & Laboratory Science

313

granulocytic precursor. Thus, we are able to estimatethe proportions of normal erythroid and granulo-cytic precursors and lymphocytes in the cases ofAML, ALL, and IR.

As noted, we found that the cells on scattergramby FSC and SSC were clustered in one populationin acute leukemia cases. The geometric mean valueof SSC, being significantly greater in AML than inALL, was useful to differentiate AML from ALL.Thus this simple technique has potential applicationin clinical flow cytometry laboratories for bonemarrow immunophenotyping and analysis.

Our results show that analysis of bone marrowaspirates using an automated hematology analyzerprovides objective data about the number ofnucleated elements, and that it is valuable in thedetermination of bone marrow cellularity. ThisCD45/PI gating system is superior to the traditionalCD45/SS gating because the debris are effectivelyseparated and the cell clusters are more easilydistinguished as rectangular, not round, or irregularlyshaped. The secondary gating by FSC/SSC cancompensate for the minor error in the CD45/PIgating. This system also can be applied to leukemiacases. Therefore, we are able to estimate thepercentages of blasts and normal hematopoieticprecursors. It provides a method that allows anyclinical flow cytometry laboratory to automate bonemarrow differentials and it can also be used to assistwith the diagnosis of bone marrow pathology,including acute leukemias.

References

1. Terstappen LW, Safford M, Loken MR. Flow cytometricanalysis of human bone marrow III. Neutrophil matur-ation. Leukemia 1990;4:657-663.

2. Sakamoto C, Yamane T, Ohta K, Hino M, Tsuda I,Tatsumi N. Automated enumeration of cellular compo-

sition in bone marrow aspirate with the CELL-DYN4000automated hematology analyzer. Acta Haematol 1999;101:130-134.

3. Fujimoto H, Sakata T, Hamaguchi Y, Shiga S, TohyamaK, Ichiyama S, Wang F, Houwen B. Flow cytometricmethod for enumeration and classification of reactiveimmature granulocyte population. Cytometry 2000;42:371-378.

4. Poon AO, McClure PD, Atkinson JB, Ing G. Diagnosticvalue of the Hemalog D-90 in classifying acute childhoodleukaemias. Clin Lab Haemat 1981;3:239-244.

5. den Ottolander GJ, Baelde HA, Huibregtsen L, PaauweJL, van der Burgh JF. The H1 automated differentialcounter in determination of bone marrow remission inacute leukemia. Am J Clin Pathol 1995;103:492-495.

6. Stelzer GT, Shults KE, Loken MR. CD45 gating forroutine flow cytometric analysis of human bone marrowspecimens. Ann N Y Acad Sci 1993;677:265-280.

7. Knapp JC, Russel KTR, Ricordi C. An automated methodfor flow-cytometric analysis of human vertebral bodymarrow. Transplant proc 1997;29,1953.

8. Davey FR, Hutchison RE. Hematopoiesis. In: ClinicalDiagnosis and Management by Laboratory Methods, 20thed (Henry JB, Ed), Saunders, Philadelphia, 2001; pp 539.

9. Tatsumi J, Tatsumi Y, Tatsumi N. Counting anddifferential of bone marrow cells by an electronic method.Am J Clin Pathol 1986;86:50-54.

10. Ozkaynak MF, Scribano P, Gomperts E, Lzadi P, MillinR, Isaacs H Jr, Ettinger LJ. Comparative evaluation ofthe bone marrow by the volumetric method, particlesmears, and biopsies in pediatric disorders. Am J Hematol1988;29:144-147.

11. Rainer RO, Hodges L, Seltzer GT. CD45 gating correlateswith bone marrow differential. Cytometry 1995;22:139-145.

12. Tsuji T, Sakata T, Hamaguchi Y, Wang F, Houwen B. Newrapid flow cytometric method for the enumeration ofnucleated red blood cells. Cytometry 1999;37:291-301.

13. Hodgetts J, Hoy JG, Jacobs A. Assessment of DNAcontent and cell cycle distribution of erythoid and myeloidcells from bone marrow. J Clin Pathol 1988;41:1120-1124.

14. Darznykiewicz Z, Traganos F, Kapuscinski J, Staiano-Coico L, Melamed MR. Accessibility of DNA in situ tovarious fluorochromes: relationship to chromatin changesduring erythoid differentiation of Friend leukaemia cells.Cytometry 1984;5:355-363.

.

Automated analysis of bone marrow aspirates