Multiparameter flow cytometry (MFC) has the potential to allow for sensitive and specific monitoring of residual disease (RD) in acute myeloid leukemia (AML). The use of MFC for RD monitoring assumes that AML cells identified by their immunophenotype at diagnosis can be detected during remission and at relapse. AML cells from 136 patients were immunophenotyped by MFC at diagnosis and at first relapse using 9 panels of 3 monoclonal antibodies. Immunophenotype changes occurred in 124 patients (91%); they consisted of gains or losses of discrete leukemia cell populations resolved by MFC (42 patients) and gains or losses of antigens on leukemia cell populations present at both time points (108 patients). Antigen expression defining unusual phenotypes changed frequently: CD13, CD33, and CD34, absent at diagnosis in 3, 33, and 47 cases, respectively, were gained at relapse in 2 (67%), 15 (45%), and 17 (36%); CD56, CD19, and CD14, present at diagnosis in 5, 16, and 20 cases, were lost at relapse in 2 (40%), 6 (38%), and 8 (40%). Leukemia cell gates created in pretreatment samples using each 3-antibody panel allowed identification of relapse AML cells in only 68% to 91% of cases, but use of 8 3-antibody panels, which included antibodies to a total of 16 antigens, allowed identification of relapse AML cells in all cases. Thus, the immunophenotype of AML cells is markedly unstable; nevertheless, despite this instability, MFC has the potential to identify RD in AML if multiple antibody panels are used at all time points.

Introduction

Monitoring of residual disease (RD) has the potential to improve treatment outcome in acute myeloid leukemia (AML) by allowing therapy to be altered based on leukemia cell burden. Strategies developed for RD monitoring include the use of multiparameter flow cytometry (MFC). In MFC, cells are stained simultaneously with multiple antibodies labeled with different fluorochromes, allowing resolution of cells into discrete populations defined by patterns of antigen coexpression.1 By MFC, AML cells exhibit abnormal patterns of antigen coexpression, which allow them to be resolved from normal cells.2,3 MFC has the potential to provide for widely applicable, rapid, cost-effective, sensitive, and specific evaluation of RD in post-treatment specimens.3,4 

The use of MFC to monitor RD in AML assumes that leukemia cells identified by their immunophenotype at diagnosis can be detected during and after therapy. The stability or instability of the immunophenotype of AML cells should be reflected by changes in the immunophenotype at relapse, in relation to that present at diagnosis, in patients who relapse. Because MFC resolves AML cells from normal cells based on patterns of antigen coexpression, it is ideal for immunophenotyping AML cells in relapse samples, where they commonly coexist with large numbers of normal cells. The immunophenotype is known to remain constant over time in most cases of acute lymphocytic leukemia (ALL),5 but little has been reported about immunophenotype changes in AML at relapse. Specifically, there are no large series comparing relapse and pretreatment immunophenotypes in AML studied by MFC.

Cells from 136 adult de novo AML patients enrolled in the Cancer and Leukemia Group B (CALGB) protocol 8361, “Immunophenotyping Studies in AML,” were immunophenotyped by MFC at diagnosis and at first relapse. Each case was evaluated for changes in the immunophenotype of AML cells reflecting either losses or gains of discrete populations of leukemia cells (“population changes”), losses or gains of antigens expressed on populations of leukemia cells that were present at both time points (“antigen changes”), or both.

Patients, materials, and methods

Patients

All newly diagnosed adult patients with de novo AML who were enrolled in CALGB 8361, “Immunophenotyping Studies in Acute Myeloid Leukemia,” between March 1991, when CALGB immunophenotyping began to be performed by MFC, and October 1996, and who had both pretreatment samples and samples from first relapse submitted for immunophenotyping were included in the analysis. Patients received remission induction and postremission therapy on 5 different CALGB treatment protocols, CALGB 8923, 9022, 9191, 9222, and 9420.6-10 

Morphologic studies

The diagnosis of AML and the assignment of French-American-British group (FAB) subtypes were based on standard morphologic and cytochemical criteria.11,12 All cases were centrally reviewed. Criteria for remission and for relapse were according to the National Cancer Institute-sponsored workshop on definitions of diagnosis and response in AML.13 

Cytogenetic analysis

Cytogenetic analysis was performed as part of a prospective karyotyping study, CALGB 8461, “Cytogenetic Studies in Acute Leukemia.” Bone marrow samples were processed for cytogenetic analysis by standard techniques, using direct and short-term (24-48 hours) unstimulated cultures. Chromosomes were G- or Q-banded. Karyotypes were designated according to the ISCN nomenclature.14 Two karyotypes from each clone were centrally reviewed.

Karyotypes at relapse were compared to those at diagnosis in cases successfully karyotyped at both time points. Changes were categorized according to the following definitions: same, karyotype was the same before treatment and at relapse; normal to abnormal, pretreatment karyotype was normal and relapse karyotype was abnormal; abnormal to normal, pretreatment karyotype was abnormal and relapse karyotype was normal; clonal evolution, karyotype was more complex at relapse than at diagnosis; clonal regression, karyotype was less complex at relapse than at diagnosis; clonal evolution and regression, some pretreatment abnormalities disappeared at relapse and new abnormalities appeared; new clone, the abnormal karyotype at relapse exhibited totally different features from that at diagnosis.

MFC

Multiparameter flow cytometry was performed in the Laboratory of Flow Cytometry at Roswell Park Cancer Institute, as previously described.15,16 Samples were sent to the laboratory by overnight mail at ambient temperature in tubes containing sodium heparin. Marrow or blood (2 mL) was filtered through a 75-μm mesh. Cells were pelleted, washed twice with 15 mL phosphate-buffered saline (PBS) with 10 U/mL heparin in the first wash, resuspended in 1 mL PBS, and then incubated with 70 μL of a 3 mg/mL solution of mouse immunoglobulin (IgG fraction) on ice for 10 minutes. Cell suspension (50 μL) was added to tubes containing panels of fluoresceinated, phycoerythrinated and biotinylated or phycoerythrin-CY5-conjugated monoclonal antibodies to CD45, CD14, and HLADr; CD3, CD4, and CD8; CD15, CD34, and CD56; CD33, CD13, and HLADr; CD16, CD32, and CD64; CD7, CD13, and CD2; CD11b, CD13, and CD33; CD38, CD34, and HLADr; and CD33, CD13, and CD19 (Becton Dickinson, San Jose, CA; Caltag, South San Francisco, CA; and Coulter, Hialeah, FL), as well as to tubes containing either isotype controls or no antibodies. Cells were incubated on ice for 15 minutes, then washed in 3.5 mL of an ammonium chloride solution to remove excess antibody and to lyse red blood cells. When biotinylated antibodies were used, tubes were centrifuged at 1500g for 3 minutes, decanted, blotted, and incubated with 10 μL phycoerythrin-Texas red avidin (Southern Biotech, Birmingham, AL) or Red 670 (BRL-Gibco, Rockville, MD) on ice for 15 minutes. All samples were washed with 3.5 mL PBS and fixed overnight in 0.5 mL 2% ultrapure formaldehyde (Polysciences, Warrington, PA).

Cell viability was determined by ethidium monoazide fluorescence, as previously described.17 All specimens exhibited more than 98% viable cells after processing and were thus adequate for analysis according to the National Committee for Clinical Laboratory Standards guidelines.18 Forward and side scatter and 3 colors of fluorescence were measured for each sample on a FACScan flow cytometer (Becton Dickinson Immunocytometry Systems, San Jose, CA); 20 000 events were analyzed in each sample. Data were analyzed using WinList multiparameter analysis software (Verity Software House, Topsham, ME). To identify populations of leukemia cells in each case of AML, the 3-antibody panel was chosen that best resolved leukemia cells as dense clusters with a minimum of normal cell contamination. Regions were drawn around the abnormal cell populations and the data were reanalyzed (backgated) to produce bivariate displays in the forward- versus side-scatter plot. New regions were drawn around the cells in this display (“leukemia gate”), and these scatter regions were then used to analyze all other panels. Cases were called positive for an antigen if it was expressed on 10% or more of cells in the leukemia gate and if expression on the surface of leukemia cells was confirmed by visual analysis of bivariate histograms, regardless of the intensity of antigen expression (dim, bright, or intermediate).

Antigen expression on AML cells is frequently heterogeneous, and patterns of antigen expression define one or more distinct clusters of cells in each case of AML. We use the term “population” to refer to a discrete cluster of cells defined by the expression of a unique set of antigens that are resolved within the most informative bivariate views. Multiple discrete clusters of cells, even if connected by an apparent low density of cells between them, represent multiple populations. In this way, the loss or gain of an entire population can be determined by comparing the patterns produced between the pretreatment and relapse specimens. For example, in Figure1A, at diagnosis a single population was present that expressed CD15 and CD11b homogeneously and CD13 heterogeneously, and did not express CD34. At relapse, there is complete absence of CD15 and CD11b expression; there is homogeneous expression of CD13 and heterogeneous expression of CD34. The populations present at diagnosis and at relapse are different; there is complete loss of the original population and appearance of a different population. The population present at relapse exhibits a pattern of antigen expression that differs completely from that of the population present at diagnosis.

Fig. 1.

Examples of immunophenotype changes at relapse, compared to diagnosis.

Bivariate displays of forward and side scatter (FSC, SSC) and of coexpression of 2 antigens are shown. Leukemia gates are shown in the forward- versus side-scatter displays. Lines in bivariate displays of antigen coexpression are positioned to define positive and negative regions based on binding of isotype controls. Panel A illustrates a population change at relapse. At diagnosis a single population was present that expressed CD15 and CD11b homogeneously and CD13 heterogeneously, and did not express CD34. At relapse, there was homogeneous expression of CD13 and heterogeneous expression of CD34, with complete absence of CD15 and CD11b expression. The populations present at diagnosis and at relapse are different; there is loss of the original population and appearance of a different population. Panel B illustrates antigen change at relapse. Gain of CD2 and loss of CD56 are seen at relapse on a population with bright CD34 expression, present both at diagnosis and at relapse.

Fig. 1.

Examples of immunophenotype changes at relapse, compared to diagnosis.

Bivariate displays of forward and side scatter (FSC, SSC) and of coexpression of 2 antigens are shown. Leukemia gates are shown in the forward- versus side-scatter displays. Lines in bivariate displays of antigen coexpression are positioned to define positive and negative regions based on binding of isotype controls. Panel A illustrates a population change at relapse. At diagnosis a single population was present that expressed CD15 and CD11b homogeneously and CD13 heterogeneously, and did not express CD34. At relapse, there was homogeneous expression of CD13 and heterogeneous expression of CD34, with complete absence of CD15 and CD11b expression. The populations present at diagnosis and at relapse are different; there is loss of the original population and appearance of a different population. Panel B illustrates antigen change at relapse. Gain of CD2 and loss of CD56 are seen at relapse on a population with bright CD34 expression, present both at diagnosis and at relapse.

The antigens present on each population of leukemia cells were identified by inspection of all 3 bivariate views of all of the 3-antibody combinations. Because the position of a given population can be followed by the expression of antigens on it that are unchanged, those antigens that are lost or gained by a specific population can be identified. Clearly, if all of the antigens associated with a population are lost, then the population is lost. Similarly, a new population can be demonstrated to have arisen at relapse if antigens are acquired that together define a population distinct from the populations that were present at diagnosis. The number of antigens defining a population is variable.

All cases were analyzed for gains and losses of leukemia cell populations at relapse and were classified as having a “population change” if at least one population was gained or lost between the 2 time points. Cases in which at least one common population was present at diagnosis and at relapse were also analyzed for losses and gains of antigen expression on the population(s) common to both time points, and were classified as having an “antigen change” if at least one antigen was lost or gained. Antigen loss or gain was defined by conversion from antigen positivity to antigen negativity, or vice versa, as defined above, and by clear evidence of loss or gain by visual comparison of bivariate histograms. Cases in which there was no common population at diagnosis and at relapse were classified as having a population change and as being unevaluable for antigen change.

Statistical analysis

Frequencies of antigen expression at diagnosis and at relapse were compared using the McNemar test for dependent proportions.19 The association between risk factors and immunophenotype change groups was analyzed using exact tests for contingency tables.19 Disease-free survival (DFS) was measured from the time of attainment of complete response (CR) to the time of relapse. The distribution of DFS times was calculated by the Kaplan-Meier method,20 and differences between subgroups were tested with the log-rank statistic.21 

Results

A total of 153 adult patients with de novo AML enrolled in CALGB 8361 between March 1991 and October 1996 had bone marrow specimens submitted for immunophenotyping by MFC both at diagnosis and at first relapse. Pretreatment or relapse specimens from 15 patients were unable to be evaluated; reasons included delayed receipt (6 patients), staining with the wrong antibody panels (2 patients), and failure to identify leukemia cells by MFC (7 patients). Two additional patients were excluded from the analysis because their diagnoses were changed from AML to myelodysplastic syndrome following central review of pretreatment bone marrow morphology. The remaining 136 eligible and evaluable patients were included in the analysis.

Clinical characteristics of the 136 AML patients successfully immunophenotyped at diagnosis and at first relapse are shown in Table1. Median age was 47 years. Median percentage of bone marrow blasts was 74%. The most common FAB types were M2, M4, M1, M3, and M5. DFS ranged from 1 month to 4.7 years; median DFS was 8.6 months in this series that included only patients who relapsed.

Table 1.

Pretreatment characteristics and treatment

Characteristic  
Age (y)  
 Median 47 
 Range 17-81  
Sex [no. (%)]  
 Male 73 (54)  
 Female 63 (46)  
WBC (×109/L)  
 Median 18.6 
 Range 0.6-281.0  
BM blasts (%)  
 Median 74  
 Range 7-99  
FAB [no. (%)]  
 M0 3 (2)  
 M1 17 (13)  
 M2 49 (36) 
 M3 16 (12)  
 M4 30 (22)  
 M5 14 (10) 
 M6 2 (1)  
 Unclassified 5 (4)  
CALGB treatment protocol [no. (%)]  
 8923 31 (23)  
 9022 15 (11) 
 9191 14 (10)  
 9222 71 (52)  
 9420 5 (4) 
DFS  
 Median 8.6 mo 
 Range 1 mo-4.7 y 
Characteristic  
Age (y)  
 Median 47 
 Range 17-81  
Sex [no. (%)]  
 Male 73 (54)  
 Female 63 (46)  
WBC (×109/L)  
 Median 18.6 
 Range 0.6-281.0  
BM blasts (%)  
 Median 74  
 Range 7-99  
FAB [no. (%)]  
 M0 3 (2)  
 M1 17 (13)  
 M2 49 (36) 
 M3 16 (12)  
 M4 30 (22)  
 M5 14 (10) 
 M6 2 (1)  
 Unclassified 5 (4)  
CALGB treatment protocol [no. (%)]  
 8923 31 (23)  
 9022 15 (11) 
 9191 14 (10)  
 9222 71 (52)  
 9420 5 (4) 
DFS  
 Median 8.6 mo 
 Range 1 mo-4.7 y 

WBC indicates white blood cells; BM, bone marrow; CALGB, Cancer and Leukemia Group B; DFS, disease-free survival.

Immunophenotypes changed at relapse in 124 of 136 patients (91%) (Table 2). Immunophenotype changes were of 2 kinds: gain or loss of discrete populations of leukemia cells resolved by MFC (“population changes”), and gain or loss of antigens on populations of leukemia cells present both at diagnosis and at relapse (“antigen changes”). Of the 124 cases of AML with immunophenotype changes at relapse, 16 had population changes alone, 82 had antigen changes alone, and 26 had population changes as well as antigen changes on populations present both at diagnosis and at relapse. Leukemia cell immunophenotypes were identical at diagnosis and at relapse in only 12 of the 136 patients studied at both time points (9%). Of note, all antigens were not studied in all cases (Table3), and it is therefore possible that the frequency of immunophenotype changes at relapse was slightly underestimated. There were also changes in both size (forward scatter) and granularity (side scatter) of AML cells at relapse; both increases and decreases occurred in both parameters.

Table 2.

Changes in immunophenotype at relapse, compared to diagnosis

 Number* Percent  
Immunophenotype change 124 91  
 Antigen change only 82 60 
 Population change only 16 12  
 Both antigen and population changes 26 19  
No change 12 
 Number* Percent  
Immunophenotype change 124 91  
 Antigen change only 82 60 
 Population change only 16 12  
 Both antigen and population changes 26 19  
No change 12 
*

N = 136.

In 4 cases, the original population was lost and a different population was gained.

Table 3.

Frequencies of antigen expression at diagnosis and relapse and of antigen loss and gain at relapse

Antigen N3-150 Expressed at diagnosis Lost at relapse3-151 Gained at relapse3-152 Expressed at relapse P3-153 
CD2 126 14 (11) 2 (14) 16 (14) 28 (22) .001 
CD3 129 3 (2) 3 (100) 0 (0) 0 (0) —   
CD4dim 129 93 (72) 5 (5) 13 (36) 101 (78) .10  
CD7 126 28 (22) 3 (11) 13 (13) 38 (30) .02  
CD8 128 1 (1) 1 (100) 6 (5) 6 (5) .13  
CD11b 126 52 (41) 14 (27) 9 (12) 47 (37) .21  
CD13 132 129 (98) 3 (2) 2 (67) 128 (97) .38  
CD14 129 16 (12) 6 (38) 0 (0) 10 (8) < .001 
CD15 127 104 (82) 11 (11) 6 (26) 99 (78) .14  
CD16 122 3 (2) 2 (67) 3 (3) 4 (3) 1.00  
CD19 98 5 (5) 2 (40) 5 (5) 8 (8) .45  
CD32 122 93 (76) 11 (12) 8 (28) 90 (74) .36  
CD33 130 97 (75) 10 (10) 15 (45) 102 (78) .42  
CD34 129 82 (64) 3 (4) 17 (36) 96 (74) .003 
CD38 83 80 (96) 3 (4) 1 (33) 78 (94) .13  
CD56 121 20 (17) 8 (40) 14 (14) 26 (21) .29  
CD64 122 61 (50) 16 (26) 10 (16) 55 (45) .17  
HLADr 132 100 (76) 6 (6) 5 (16) 99 (74) .55  
Antigen N3-150 Expressed at diagnosis Lost at relapse3-151 Gained at relapse3-152 Expressed at relapse P3-153 
CD2 126 14 (11) 2 (14) 16 (14) 28 (22) .001 
CD3 129 3 (2) 3 (100) 0 (0) 0 (0) —   
CD4dim 129 93 (72) 5 (5) 13 (36) 101 (78) .10  
CD7 126 28 (22) 3 (11) 13 (13) 38 (30) .02  
CD8 128 1 (1) 1 (100) 6 (5) 6 (5) .13  
CD11b 126 52 (41) 14 (27) 9 (12) 47 (37) .21  
CD13 132 129 (98) 3 (2) 2 (67) 128 (97) .38  
CD14 129 16 (12) 6 (38) 0 (0) 10 (8) < .001 
CD15 127 104 (82) 11 (11) 6 (26) 99 (78) .14  
CD16 122 3 (2) 2 (67) 3 (3) 4 (3) 1.00  
CD19 98 5 (5) 2 (40) 5 (5) 8 (8) .45  
CD32 122 93 (76) 11 (12) 8 (28) 90 (74) .36  
CD33 130 97 (75) 10 (10) 15 (45) 102 (78) .42  
CD34 129 82 (64) 3 (4) 17 (36) 96 (74) .003 
CD38 83 80 (96) 3 (4) 1 (33) 78 (94) .13  
CD56 121 20 (17) 8 (40) 14 (14) 26 (21) .29  
CD64 122 61 (50) 16 (26) 10 (16) 55 (45) .17  
HLADr 132 100 (76) 6 (6) 5 (16) 99 (74) .55  
F3-150

Number of cases studied for the antigen on the same population both at diagnosis and at relapse.

F3-151

Number (percentage) of cases with antigen expression at diagnosis which lost expression of the antigen at relapse.

F3-152

Number (percentage) of cases without antigen expression at diagnosis which gained expression of the antigen at relapse.

F3-153

McNemar test of proportions for paired data.

Population changes occurred in a total of 42 patients (31%). Ninety-seven patients (71%) had a single population of leukemia cells at diagnosis; 12 of these 97 had population changes at relapse, including 8 who gained a second population, and 4 who relapsed with a single population that was different from the one that had been present at diagnosis (Figure 1A). Thirty-nine patients (28%) had 2 discrete populations of leukemia cells at diagnosis; 30 of them relapsed with only 1 of the 2 populations.

Antigen changes occurred in 108 of the 132 cases with at least one preserved population (82%). Percentages of cases expressing each antigen pretreatment and at relapse are shown in Table 3. At relapse, a higher percentage of cases expressed CD2 (22% versus 11%,P = .001), CD34 (74% versus 64%, P = .003), and CD7 (30% versus 22%, P = .02), and a lower percentage expressed CD14 (8% versus 12%, P < .001). CD33 was expressed in 75% of cases at diagnosis and 78% at relapse (P = .42).

Frequencies of loss and gain of each antigen at relapse are also shown in Table 3. Changes in antigen expression present in at least 15 cases included gain of CD34 (17 patients), loss of CD64 and gain of CD2 (16 patients each), and gain of CD33 (15 patients). Antigen expression defining unusual phenotypes at diagnosis frequently changed at relapse (Table 3). CD13, CD33, and CD34, absent at diagnosis in 3, 33, and 47 patients, respectively, were gained at relapse in 2 (67%), 15 (45%), and 17 (36%), and CD56, CD19, and CD14, present at diagnosis in 5, 16, and 20 patients, were lost at relapse in 2 (40%), 6 (38%), and 8 (40%). An example of gain of CD2 and loss of CD56 at relapse is shown in Figure 1B.

Results of cytogenetic analysis of both pretreatment and relapse bone marrow samples were available for 72 patients (Table4). Karyotype changes were observed in 40 (56%); this frequency is lower than the frequency of immunophenotype changes. Immunophenotype changes were present without karyotype changes in 29 patients, whereas only 4 patients had karyotype changes without immunophenotype changes. There was no apparent correlation between type of immunophenotype change and type of karyotype change.

Table 4.

Association between immunophenotype and karyotype changes at relapse

Karyotype change Immunophenotype change 
Population Antigen Both None 
None 32 19 3  
Abnormal to normal 2  
Clonal evolution 19 14 1  
Clonal regression 0  
New clone 0  
Normal to abnormal 0  
Evolution and regression 
Total 72 46 15 
Karyotype change Immunophenotype change 
Population Antigen Both None 
None 32 19 3  
Abnormal to normal 2  
Clonal evolution 19 14 1  
Clonal regression 0  
New clone 0  
Normal to abnormal 0  
Evolution and regression 
Total 72 46 15 

The distribution of age and pretreatment cytogenetics among the population and antigen change groups is shown in Table5. Patients' ages were similar in the different groups, and there were no differences in distribution of pretreatment karyotypes that might be associated with differences in DFS. Of note, population changes at relapse were more frequent in patients with normal karyotypes than in those with chromosomal abnormalities (favorable and other) (P = .03), suggesting that AML with specific chromosomal abnormalities generally has stable cell populations, whereas cases with normal karyotypes more frequently exhibit populations changes at relapse.

Table 5.

Distribution of age and cytogenetics among immunophenotype change groups

Immunophenotype group  Age P Cytogenetics P 
Younger than 60
n = 96 
At least 60
n = 40 
Favorable5-150
n = 14 
Normal
n = 49 
Other
n = 45  
Population change 30 33 .84 14 39 20 .075-151 
No population change 70 68  86 61 80  
Antigen change 82 81 1.00 71 91 82 .14  
No antigen change 18 19  29 18  
Population + antigen change 20 18 .82 14 27 11 .16  
No population + antigen change 80 83  86 73 89  
Immunophenotype group  Age P Cytogenetics P 
Younger than 60
n = 96 
At least 60
n = 40 
Favorable5-150
n = 14 
Normal
n = 49 
Other
n = 45  
Population change 30 33 .84 14 39 20 .075-151 
No population change 70 68  86 61 80  
Antigen change 82 81 1.00 71 91 82 .14  
No antigen change 18 19  29 18  
Population + antigen change 20 18 .82 14 27 11 .16  
No population + antigen change 80 83  86 73 89  

Entries (except P values) are percentages of patients.

F5-150

t(8;21) and inv(16).

F5-151

Abnormal (favorable + other) versus normal:P = .03.

Median DFS for all 136 patients was 8.6 months (95% confidence interval, 7.7-9.1 months). DFS did not differ significantly between patients with and without antigen changes (medians 8.4 versus 8.4 months; P = .39), population changes (medians 8.4 versus 8.9 months; P = .87), or any change in immunophenotype (medians 8.4 versus 8.9 months; P = .89) at relapse.

Four patients relapsed more than 3 years from attaining CR, at 3.4, 3.5, 4.1, and 4.7 years. Two had both population and antigen changes, one had antigen changes, and one lost the original population and gained a new one. Thus immunophenotype changes were present at relapse in all 4 patients and included population changes in 3 of the 4.

We sought to determine whether, despite the very high frequency of immunophenotype changes, leukemia cell gates created in pretreatment samples would identify AML cells at relapse. Leukemia cell gates created in pretreatment samples using each 3-antibody panel allowed identification of relapse AML cells in only 68% to 91% of cases, but use of 8 3-antibody panels, which included antibodies to a total of 16 antigens, allowed identification of relapse AML cells in all cases (Table 6). Each case was unique due to specific antigen changes, so that relapse could not be identified in all cases with fewer than 8 antibody panels.

Table 6.

Frequency of identification of acute myeloid leukemia cells at relapse using leukemia cell gates created in pretreatment samples

Panel Antigens % cases with AML cells identified at relapse 
CD45, CD14, HLADr 87  
CD3, CD4, CD8 86  
CD15, CD34, CD56 73  
CD33, CD13, HLADr 79  
CD16, CD32, CD64 79  
CD7, CD13, CD2 88  
CD11b, CD13, CD33 77  
CD38, CD34, HLADr 68  
CD33, CD13, CD19 91  
1, 3-9 CD45, CD14, HLADr, CD15, CD34, CD56, CD33, CD13, CD16, CD32, CD64, CD7, CD2, CD11b, CD38, CD19 100 
Panel Antigens % cases with AML cells identified at relapse 
CD45, CD14, HLADr 87  
CD3, CD4, CD8 86  
CD15, CD34, CD56 73  
CD33, CD13, HLADr 79  
CD16, CD32, CD64 79  
CD7, CD13, CD2 88  
CD11b, CD13, CD33 77  
CD38, CD34, HLADr 68  
CD33, CD13, CD19 91  
1, 3-9 CD45, CD14, HLADr, CD15, CD34, CD56, CD33, CD13, CD16, CD32, CD64, CD7, CD2, CD11b, CD38, CD19 100 

AML indicates acute myeloid leukemia.

Discussion

The use of MFC to monitor RD during morphologic remission and to predict relapse depends on the presence of leukemia cells with phenotypes different from those of normal cells and on their detection throughout the course of a patient's leukemia. We studied AML cells from 136 patients at diagnosis and at relapse by MFC, using 9 panels of 3 monoclonal antibodies. We demonstrated immunophenotype changes in leukemia cells at relapse, compared to diagnosis, in 124 of the 136 patients (91%). Changes included both gains and losses of populations of AML cells, and gains and losses of antigens on populations present at both time points. This very high frequency of immunophenotype changes at relapse has potential implications for detection of RD by MFC in AML, in that loss of sensitivity for detecting residual cells in follow-up specimens will occur if leukemia cells fall outside of the criteria established at diagnosis and are therefore lost to detection. We indeed found that no single antibody panel nor group of antibody panels could be consistently used to identify AML cells at relapse. However, when 8 3-antibody panels including antibodies to 16 antigens were used together, AML cells were identified at relapse in all cases. Thus, because of the high frequency of changes in the immunophenotype of AML cells during the course of the disease, RD detection by MFC is likely to require ongoing monitoring with multiple panels of 3 or 4 antibodies.

The presence of aberrant phenotypes in AML is well documented. Terstappen and coworkers reported aberrant antigen expression at diagnosis in 80 cases of AML, including expression of nonmyeloid antigens, asynchronous expression of myeloid antigens, overexpression of myeloid antigens, and absence of expression of myeloid antigens.2 Reading and colleagues reported unusual antigen coexpression patterns (combinations present on ≤ 0.1% of cells in normal marrow) on 10% or more of blasts in 232 of 272 patients with AML (85%).3 Asynchronous expression of myeloid antigens occurred in 70%, and coexpression of T-lymphoid, B-lymphoid, or natural killer antigens with myeloid antigens in 38%, 13%, and 21% of cases, respectively. Macedo and associates found at least one aberrant phenotype in 29 of 40 AML patients, including 25 with asynchronous antigen expression, 15 with coexpression of lymphoid antigens, 13 with abnormal forward/side scatter (size versus granularity) distribution, and 7 with antigen overexpression.22 Most AML patients with leukemia-associated phenotypes in that series had more than one abnormality. The frequent presence of more than one subset of AML cells, as reported here, has also been previously described.22,23 

Little has been reported to date about the stability or instability of AML cell immunophenotypes over time in individual patients. Using single-color indirect immunofluorescence and flow cytometry, Thomas and coworkers demonstrated increased percentages of CD34+ and CD33+ cells and decreased percentages of CD13+and CD15+ cells in samples from 66 adult AML patients at relapse.24 These findings were interpreted as being consistent with a loss of myeloid differentiation. In contrast, we found frequent gain of CD34 without changes in CD33, CD13, or CD15 expression, consistent with aberrant differentiation rather than loss of myeloid differentiation. Three small series have looked at immunophenotype changes by MFC in the context of RD detection. Using 2-color flow cytometry with 10 to 12 antibodies, Drach and colleagues25 found preservation of leukemia-associated phenotypes in 6 of 7 patients with AML at relapse. Similarly, Reading and coworkers3 found that, despite variation in marker expression, “unusual phenotypes” persisted at relapse in 12 of 13 patients. Macedo and associates26 studied diagnosis and relapse samples from 16 patients by 3-color flow cytometry; changes in expression of one or more antigens were found in 10. We have previously reported loss of CD56 expression at relapse in 2 of 5 patients with t(8;21)(q22;q22) with CD56 expression at diagnosis,15 and gain or loss of the lymphoid antigens CD2, CD7, or CD56 in 5 of 7 AML patients with 11q23 translocations.16 The present report represents the first large series looking at immunophenotype changes in AML by MFC. We found a very high frequency of changes in immunophenotype, including loss of infrequently expressed antigens (CD56, CD19, and CD14) when these had been expressed at diagnosis, and gain of frequently expressed antigens (CD13, CD33, and CD34) when these had not been expressed. The patient population studied here consisted of de novo AML patients with a relatively low median age (47 years) and a relatively high median white blood cell count (18.6 × 109/L); it is not known to what extent the findings can be generalized to other groups such as older patients and patients with secondary AML.

Multiparameter flow cytometry has been used successfully to detect RD in AML, albeit with significant rates of both false-negative and false-positive results.3,4 In a recent extensive review of RD detection by MFC,27 Campana and Coustan-Smith recognized immunophenotypic shifts as a cause of false-negative results, and suggested that this obstacle may be overcome by the use of multiple antibody combinations. They described the use of 12 different 4-antibody combinations, including antibodies to CD13, CD33, CD34, CD11b, CD65, CD15, CD56, CD38, CD19, CD2, CD7, CD117, and HLA-Dr, and stated that approximately 60% of cases of childhood AML can be monitored for RD with a sensitivity of 10−3 to 10−4 using this approach. Our demonstration of a very high frequency of immunophenotype changes at relapse also lends support to the concept that multiple antibody panels must be used for monitoring RD, in that the use of 8 3-antibody panels was required to identify leukemia cells at relapse in all cases in our study, and the immunophenotype changes that are present at relapse are likely to have occurred during remission. It should be noted that the analysis performed here only addressed the ability to identify leukemia cells at relapse (sensitivity) and did not address whether the criteria used to identify leukemia cells with each antibody panel allowed leukemia cells to be distinguished from normal cells (specificity). Moreover, AML cells will be more difficult to detect in remission, compared to relapse, because they are present in much smaller numbers.

The high frequency of immunophenotype changes in AML at relapse contrasts with findings in ALL. CALGB recently reported 37 adult ALL patients studied by MFC both at diagnosis and at first relapse.5 Changes in immunophenotype were less frequent and generally consisted of minor shifts in antigen expression. The most common immunophenotype change was loss or gain of a myeloid marker, rather than population changes or altered lymphoid antigen expression.

In addition to the implications for RD detection, our demonstration of a very high frequency of immunophenotype changes in AML at relapse provides compelling evidence for genetic instability in this disease. Cytogenetic changes also occur in AML at relapse, albeit less commonly than the changes in immunophenotype that we report. The higher frequency of immunophenotype change compared to karyotype change at relapse may be explained by the much larger number of cells examined by flow cytometric, compared to cytogenetic, analysis. We found cytogenetic changes in 40 of 72 AML patients (56%) at relapse in the series reported here. Estey and coworkers28 found new or additional cytogenetic changes in 131 of 212 AML patients (62%). Neither presence of cytogenetic change nor type of cytogenetic change from diagnosis to relapse correlated with remission duration in Estey's study. We also found a lack of correlation between immunophenotype changes and DFS in our study.

One of the most common antigen changes in our series, as in the study by Thomas and colleagues,24 was gain of expression of the stem cell antigen CD34. Expression of CD34 on AML cells is associated with expression of the multidrug resistance protein p-glycoprotein (Pgp), an energy-dependent drug efflux pump encoded by the MDR1 gene,29,30 which is expressed in normal hematopoietic stem cells.31 Expression of Pgp has been reported to increase at relapse of AML, compared to diagnosis, in some studies,32,33 but not in others.34-36Importantly, functional drug efflux in cells that express Pgp appears to be restricted to CD34+ cells.37-39 Thus increased expression of CD34 at relapse might be associated with an increase in functional drug efflux. More common expression of multidrug resistance-associated protein (MRP) may also play a role in clinical drug resistance at relapse,34,35 but expression of lung-resistance protein (LRP) does not appear to increase.33,40 

Recently developed therapies in AML include the use of humanized anti-CD33 monoclonal antibody and anti-CD33 monoclonal antibody conjugated to the antitumor antibiotic calicheamicin or the plant toxin gelonin.41-43 We found that CD33 was expressed in the majority of cases of AML both at diagnosis and at relapse; almost half of the rare cases that did not express CD33 at diagnosis acquired CD33 expression at relapse. Thus, CD33 represents an excellent target for immunotherapy in AML both at diagnosis and at relapse. Nevertheless, 10 of 97 cases of AML with CD33 expression at diagnosis in our series lost CD33 expression at relapse, indicating that patients need to be retested for CD33 expression at relapse prior to use of anti-CD33-based therapies. Furthermore, given the instability of the immunophenotype of AML cells, loss of CD33 expression might occur as a mechanism of resistance to anti-CD33 antibody-based therapies, and patients receiving these novel therapies should be monitored for the development of this mechanism of resistance.

Monitoring of RD by MFC holds promise for improving outcome in AML, but its optimal application requires definition of optimal monitoring strategies. We have demonstrated that the immunophenotype of AML cells is markedly unstable during the course of the disease, as evidenced by the presence of immunophenotype changes at relapse in almost all patients. The instability of the immunophenotype in AML is likely to mandate the ongoing use of multiple antibody panels for RD monitoring by MFC; strategies using limited numbers of antibody panels will likely lack sensitivity, resulting in false-negative data. RD monitoring by MFC using multiple 3- or 4-antibody panels is a promising technique that requires validation in prospective clinical trials.

The following institutions participated in the study: CALGB Statistical Office, Durham, NC- Stephen George, supported by CA33601; Dana Farber Cancer Institute, Boston, MA, George P. Canellos, supported by CA32291; Duke University, Durham, NC, Jeffrey Crawford, supported by CA47577; Long Island Jewish Medical Center, Lake Success, NY, Marc Citron, supported by CA11028; Medical University of South Carolina, Charleston, SC, Mark R. Green, supported by CA03927; Mount Sinai Medical Center, New York, NY, Lewis R. Silverman, supported by CA04457; North Shore University Hospital CCOP, Manhasset, NY, Vincent Vinciguerra, supported by CA35279; Rhode Island Hospital, Providence, RI, Louis A. Leone, supported by CA08025; Roswell Park Cancer Institute, Buffalo, NY, Ellis Levine, supported by CA02599; State University of New York Upstate Medical University, Syracuse, NY, Stephan L. Graziano, supported by CA21060; University of Alabama, Birmingham, Robert Diasio, supported by CA 47545; University of California at San Diego, Stephen L. Seagren, supported by CA11789; University of Chicago Medical Center, Chicago, IL, Gini Fleming, supported by CA41287; University of Iowa, Iowa City, Gerald H. Clamon, supported by CA47642; University of Maryland Cancer Center, Baltimore, David Van Echo, supported by CA31983; University of Massachusetts Medical Center, Worcester, Marc Stewart, supported by CA37135; University of Minnesota, Minneapolis, Bruce A. Peterson, supported by CA16450; University of Missouri/Ellis Fischel Cancer Center, Columbia, Michael C. Perry, supported by CA12046; University of North Carolina at Chapel Hill, Thomas C. Shea, supported by CA47559; University of Tennessee Memphis, Memphis, Harvey B. Niell, supported by CA47555; Wake Forest University School of Medicine, Winston-Salem, NC, David D. Hurd, supported by CA03927.

Supported by National Cancer Institute grants CA02599 (M.R.B., C.C.S., N.S.), CA33601 (R.K.D.), CA77658 (K.M., C.D.B.), CA03927 (B.L.P.), CA35279 (J.E.K.), CA47577 (J.O.M.), CA32291 (R.M.S.), CA47559 (A.J.C.), and CA41287 (R.A.L.). The research for CALGB 8361 was supported, in part, by grants from the National Cancer Institute (CA31946) to the Cancer and Leukemia Group B (Richard L. Schilsky, chairman). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute. Participating institutions are listed in the.

The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 U.S.C. section 1734.

References

References
1
Stewart
CC
Clinical applications of flow cytometry: immunologic methods for measuring cell membrane and cytoplasmic antigens.
Cancer.
69(suppl 1)
1992
1543
1552
2
Terstappen
LWMM
Safford
M
Könemann
S
et al. 
Flow cytometric characterization of acute myeloid leukemia, II: phenotypic heterogeneity at diagnosis.
Leukemia.
6
1992
70
80
3
Reading
CL
Estey
EH
Huh
YO
et al. 
Expression of unusual immunophenotype combinations in acute myelogenous leukemia.
Blood.
81
1993
3083
3090
4
San Miguel
JF
Martı́nez
A
Macedo
A
et al. 
Immunophenotyping investigation of minimal residual disease is a useful approach for predicting relapse in acute myeloid leukemia patients.
Blood.
90
1997
2465
2470
5
Czuczman
MS
Dodge
RK
Stewart
CC
et al. 
Value of immunophenotype in intensively treated adult acute lymphoblastic leukemia: Cancer and Leukemia Group B Study 8364.
Blood.
93
1999
3931
3939
6
Stone
RM
Berg
DT
George
SL
et al. 
Granulocyte-macrophage colony-stimulating factor after initial chemotherapy for elderly patients with primary acute myelogenous leukemia.
N Engl J Med.
332
1995
1671
1677
7
Moore
JO
Dodge
RK
Amrein
PC
et al. 
Granulocyte-colony stimulating factor (filgrastim) accelerates granulocyte recovery after intensive postremission chemotherapy for acute myeloid leukemia with aziridinyl benzoquinone and mitoxantrone: Cancer and Leukemia Group B Study 9022.
Blood.
89
1997
780
788
8
Tallman
MS
Andersen
JW
Schiffer
CA
et al. 
All-trans-retinoic acid in acute promyelocytic leukemia.
N Engl J Med.
337
1997
1021
1028
9
Moore
JO
Powell
B
Velez-Garcia
E
et al. 
A comparison of sequential non-cross-resistant therapy or Ara-C consolidation following complete remission in adult patients <60 years with acute myeloid leukemia: CALGB 9222.
Proc Am Soc Clin Oncol.
16
1997
14a
10
Lee
EJ
George
SL
Caligiuri
M
et al. 
Parallel phase I studies of daunorubicin given with cytarabine and etoposide with or without the multidrug resistance modulator PSC-833 in previously untreated patients 60 years of age or older with acute myeloid leukemia: results of Cancer and Leukemia Group B Study 9420.
J Clin Oncol.
17
1999
2831
2839
11
Bennett
JM
Catovsky
D
Daniel
MT
et al. 
Proposed revised criteria for the classification of acute myeloid leukemia: a report of the French-American-British Cooperative Group.
Ann Intern Med.
103
1985
620
625
12
Bennett
JM
Catovsky
D
Daniel
MT
et al. 
Proposal for the recognition of minimally differentiated acute myeloid leukaemia (AML-MO).
Br J Haematol.
78
1991
325
329
13
Cheson
BD
Cassileth
PA
Head
DR
et al. 
Report of the National Cancer Institute-sponsored workshop on definitions of diagnosis and response in acute myeloid leukemia.
J Clin Oncol.
8
1990
813
819
14
Mitelman
F
ISCN: an International System for Human Cytogenetic Nomenclature.
1995
Karger
Basel, Switzerland
15
Baer
MR
Stewart
CC
Lawrence
D
et al. 
Expression of the neural cell adhesion molecule CD56 is associated with short remission duration and survival in acute myeloid leukemia with t(8;21)(q22;q22).
Blood.
90
1997
1643
1648
16
Baer
MR
Stewart
CC
Lawrence
D
et al. 
Acute myeloid leukemia with 11q23 translocations: myelomonocytic immunophenotype by multiparameter flow cytometry.
Leukemia.
12
1998
317
325
17
Riedy
MC
Muirhead
KA
Jensen
CP
Stewart
CC
Use of a photolabeling technique to identify nonviable cells in fixed homologous or heterologous cell populations.
Cytometry.
12
1991
133
139
18
National Committee for Clinical Laboratory Standards
Clinical applications of flow cytometry: immunophenotyping of leukemic cells; proposed guidelines. Document H43-P
1993
Author
Villanova, PA
19
Agresti
A
Categorical Data Analysis.
1990
Wiley
New York, NY
20
Kaplan
EL
Meier
P
Nonparametric estimation from incomplete observations.
J Am Stat Assoc.
53
1958
475
481
21
Peto
R
Pike
MC
Armitage
P
et al. 
Design and analysis of randomized clinical trials requiring prolonged observation of each patient, II: analysis and examples.
Br J Cancer.
35
1977
1
39
22
Macedo
A
Orfao
A
Vidriales
MB
et al. 
Characterization of aberrant phenotypes in acute myeloblastic leukemia.
Ann Hematol.
70
1995
189
194
23
Macedo
A
Orfao
A
Gonzalez
M
et al. 
Immunological detection of blast cell subpopulations in acute myeloblastic leukemia at diagnosis: implications for minimal residual disease studies.
Leukemia.
9
1995
993
998
24
Thomas
X
Campos
L
Archimbaud
E
et al. 
Surface marker expression in acute myeloid leukaemia at first relapse.
Br J Haematol.
81
1992
40
44
25
Drach
J
Drach
D
Glassl
H
Gattringer
C
Huber
H
Flow cytometric determination of atypical antigen expression in acute leukemia for the study of minimal residual disease.
Cytometry.
13
1992
893
901
26
Macedo
A
San Miguel
JF
Vidriales
MB
et al. 
Phenotypic changes in acute myeloid leukaemia: implications in the detection of minimal residual disease.
J Clin Pathol.
49
1996
15
18
27
Campana
D
Coustan-Smith
E
Detection of minimal residual disease in acute leukemia by flow cytometry.
Cytometry.
38
1999
139
152
28
Estey
E
Keating
MJ
Pierce
S
Stass
S
Change in karyotype between diagnosis and first relapse in acute myelogenous leukemia.
Leukemia.
9
1995
972
976
29
Campos
L
Guyotat
D
Archimbaud
E
et al. 
Clinical significance of multidrug resistance P-glycoprotein expression on acute nonlymphoblastic leukemia cells at diagnosis.
Blood.
79
1992
473
476
30
Guerci
A
Merlin
JL
Missoum
N
et al. 
Predictive value for treatment outcome in acute myeloid leukemia of cellular daunorubicin accumulation and P-glycoprotein expression simultaneously determined by flow cytometry.
Blood.
85
1995
2147
2153
31
Chaudhary
PM
Roninson
IB
Expression and activity of P-glycoprotein, a multidrug efflux pump, in human hematopoietic stem cells.
Cell.
66
1991
85
94
32
Maslak
P
Hegewisch-Becker
S
Godfrey
L
Andreeff
M
Flow cytometric determination of the multidrug-resistant phenotype in acute leukemia.
Cytometry.
17
1994
84
93
33
Hart
SM
Ganeshaguru
K
Scheper
RJ
Prentice
HG
Hoffbrand
AV
Mehta
AB
Expression of the human major vault protein LRP in acute myeloid leukemia.
Exp Hematol.
25
1997
1227
1232
34
Hart
SM
Ganeshaguru
K
Hoffbrand
AV
Prentice
HG
Mehta
AB
Expression of the multidrug resistance-associated protein (MRP) in acute leukemia.
Leukemia.
8
1994
2163
2168
35
Schneider
E
Cowan
KH
Bader
H
et al. 
Increased expression of the multidrug resistance-associated protein gene in relapsed acute leukemia.
Blood.
85
1995
186
193
36
Zhou
D-C
Zittoun
R
Marie
J-P
Expression of multidrug resistance-associated protein (MRP) and multidrug resistance (MDR1) genes in acute myeloid leukemia.
Leukemia.
9
1995
1661
1666
37
Bailly
J-D
Muller
C
Jaffrezou
J-P
et al. 
Lack of correlation between expression and function of P-glycoprotein in acute myeloid leukemia cell lines.
Leukemia.
9
1995
799
807
38
te Boekhorst
PAW
de Leeuw
K
Schoester
M
et al. 
Predominance of functional multidrug resistance (MDR-1) phenotype in CD34+ acute myeloid leukemia cells.
Blood.
82
1993
3157
3162
39
Leith
CP
Chen
I-M
Kopecky
KJ
et al. 
Correlation of multidrug resistance (MDR1) protein expression with functional dye/drug efflux in acute myeloid leukemia by multiparameter flow cytometry: Identification of discordant MDR−/efflux+ and MDR1+/efflux− cases.
Blood.
86
1995
2329
2342
40
List
AF
Spier
CS
Grogan
TM
et al. 
Overexpression of the major vault transporter protein lung-resistance protein predicts treatment outcome in acute myeloid leukemia.
Blood.
87
1996
2464
2469
41
Caron
PC
Dumont
L
Scheinberg
DA
Supersaturating infusional humanized anti-CD33 monoclonal antibody HuM195 in myelogenous leukemia.
Clin Cancer Res.
4
1998
1421
1428
42
Pagliaro
LC
Liu
B
Munker
R
et al. 
Humanized M195 monoclonal antibody conjugated to recombinant gelonin: an anti-CD33 immunotoxin with antileukemic activity.
Clin Cancer Res.
4
1998
1971
1976
43
Sievers
EL
Appelbaum
FR
Spielberger
RT
et al. 
Selective ablation of acute myeloid leukemia using antibody-targeted chemotherapy: a phase I study of an anti-CD33 calicheamicin immunoconjugate.
Blood.
93
1999
3678
3684

Author notes

Maria R. Baer, Leukemia Section, Department of Medicine, Roswell Park Cancer Institute, Elm and Carlton Sts, Buffalo, NY 14263; e-mail: maria.baer@roswellpark.org.