Stability of leukemia-associated aberrant immunophenotypes in patients with acute myeloid leukemia between diagnosis and relapse: Comparison with cytomorphologic, cytogenetic, and molecular genetic findings
Laboratory for Leukemia Diagnostics, Department of Internal Medicine III, University Hospital Grosshadern, Ludwig-Maximilians-University, Munich, Germany
Multiparameter flow cytometry is increasingly used to monitor minimal residual disease in patients with acute myeloid leukemia to identify leukemic cells by leukemia-associated aberrant immunophenotypes (LAIPs). Changes in LAIPs during the course of the disease may be a limitation for this approach.
We analyzed 49 patients at diagnosis and relapse by flow cytometry, cytomorphology, cytogenetics, and molecular genetics.
In 37 patients (76%), at least one LAIP detectable at diagnosis was present at relapse; in 12 patients (24%), none of the original LAIPs were present in at least 1% of bone marrow cells. Three groups were identified: no change in LAIPs, partial changes in LAIPs, and complete change in LAIPs. There were significant differences across these groups with regard to changes in cytomorphology (11%, 40%, and 58% of all cases, respectively; P = 0.007), cytogenetics (15%, 20%, and 25%; not significant), and molecular genetics (18%, 0, and 86%; P = 0.002).
Based on modern therapy strategies, 50% to 80% of patients with acute myeloid leukemia (AML) achieve a complete remission (CR) (1–3). However, nearly 50% of these patients eventually relapse, which represents the major cause of treatment failure (4). Relapse occurs in patients with acute leukemia because of the persistence of a small number of leukemic cells, i.e., minimal residual disease (MRD), that are not detectable by conventional cytomorphologic techniques (5, 6). The most reliable methods for the quantification of MRD are those based on polymerase chain reaction (PCR) and multiparameter flow cytometry (MFC) (7–14). The level of MRD is a valuable tool for predicting patients' outcomes (15–20). Therefore, it is important to inquire into possible limitations of these methods. The applicability of PCR methods for MRD assessment is restricted to subgroups of AML with leukemia-specific targets that can be quantified at follow-up analysis, e.g., the fusion transcripts AML1-ETO, PML-RARA, CBFB-MYH11, and those involving MLL. However, these cases comprise only 20% of all AML cases (21). Additional genetic alterations that are targets for PCR-based detection of MRD are length mutations of the FLT3 gene (FLT3-LM) and internal partial tandem duplication of the MLL gene (MLL-PTD). These mutations occur with frequencies of 23% and 6.5%, respectively, in patients with AML and are commonly accompanied by a normal karyotype (22–24). Thus, mutation-based monitoring of MRD by PCR is applicable for up to 50% of patients with AML. The application of MFC for MRD monitoring relies on the leukemia-associated aberrant immunophenotypes (LAIPs), i.e., immunophenotypes of leukemic cells that are not or are only rarely present in normal hematopoietic cells (25, 26). Based on recent significant technical advances, a LAIP can be identified in virtually every patient with AML (16). The change of aberrant immunophenotypes during the course of disease may be a limitation for the MRD investigation by MFC and has been studied only in small series (27–34). Although the phenotypic aberrations remain constant at relapse in most cases of acute lymphoblastic leukemia (35, 36), the stability of aberrant immunophenotype in AML cases over time may be present less often. Several small series have described shifts in aberrant immunophenotype at relapse in AML cases with various frequencies, possibly related to the number of marker combinations that were applied to the respective analyses (28, 31, 37–42). In all of these studies, the detection of immunophenotype changes was focused on the subgroup of patients who displayed highly aberrant phenotypes, which comprise 60% to 80% of all AML cases only.
In this context, it is important to recognize that the relapse of AML may represent a clonal evolution of the disease or may be the occurrence of secondary AML, i.e., treatment-related AML. This is anticipated to also result in shifts of initial LAIPs at relapse (43, 44). In several previous studies, the instability of the leukemia karyotype between diagnosis and relapse has been described (45–47). However, only limited data are available that compare the incidence of immunophenotype shifts with morphologic and cytogenetic findings at relapse in patients with AML (48, 49). In the present study, we investigated the frequency of LAIP changes between diagnosis and relapse in patients with AML for whom complete morphologic, cytogenetic, and molecular genetic datasets were also available.
MATERIALS AND METHODS
We analyzed bone marrow (BM) samples from 49 adult patients with newly diagnosed AML (de novo and secondary AML) at diagnosis and at relapse. The diagnosis of AML was performed according to the French-American-British criteria and the World Health Organization classification (50, 51). Cytomorphology, cytochemistry, cytogenetics, and molecular genetics were applied in all cases as described below. The study abides by the rules of the local internal review board and the tenets of the revised Helsinki protocol (http://www.wma.net/e/policy/b3.htm).
Normal BM Samples
Normal BM samples, obtained from 26 healthy volunteers, served as controls and were analyzed by flow cytometry as described in detail below.
Multiparameter Flow Cytometry
The study was performed on mononuclear cells purified from BM aspirates by Ficoll-Hypaque density gradient centrifugation (52). After applying triple staining and isotype controls, monoclonal antibodies against 31 surface and cytoplasmic antigens were used in the following combinations designed for the detection of LAIPs (conjugated with the fluorochromes fluorescein isothiocyanate, phycoerythrin, and phycoerythrincyanin 5 (PC-5), respectively): CD34/CD2/CD33, CD7/CD33/CD34, CD34/CD56/CD33, CD11b/CD117/CD34, CD64/CD4/CD45, CD34/CD13/CD19, CD65/CD87/CD34, CD15/CD34/CD33, HLA-DR/CD33/CD34, CD4/CD13/CD14, CD34/CD135/CD117, CD34/CD116/CD33, CD90/CD117/CD34, CD34/NG2(7.1)/CD33, CD38/CD133/CD34, CD61/CD14/CD45, CD36/CD235a/CD45, CD15/CD13/CD33, CD9/CD34/CD33, CD38/CD34/CD90, CD34/CD79a/CD19, TdT/cyCD33/cyCD45, MPO/lactoferrin/cyCD15, TdT/cyCD79a/cyCD3, TdT/cyCD22/cyCD3. All antibodies were purchased from Immunotech (Marseilles, France), except for CD64 and CD15 (Medarex, Annandale, NJ, USA), CD133 (Miltenyi Biotec, Bergisch Gladbach, Germany), and myeloperoxidase (MPO) and lactoferrin (Caltag, Burlingame, CA, USA). The respective combinations of antibodies were added to 1 × 106 cells (volume 100 μl) and incubated for 10 min at room temperature. After addition of 2 ml of lysing solution (based on ammonium chloride and prepared at a local pharmaceutical institute), the samples were incubated for additional 10 min, washed twice in phosphate buffered saline, and resuspended in 0.5 ml of phosphate buffered saline. For the analysis of cytoplasmic antigens, cells were fixed and permeabilized before staining with Fix&Perm (Caltag). MFC analysis was performed with a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA, USA). Forward and right-angle light scatters were collected in addition to three-color antibody combinations to generate five characteristics per cellular event. For samples at the time of diagnosis, 20,000 events were acquired; for normal BM samples, 250,000 events were acquired, which is the number for acquisition of MRD samples. For the relapse BM samples, the same combinations of antibodies were applied and 20,000 cells/tube were collected. If the number of cells in the relapse sample was limited, the protocol for MRD samples was used, i.e., only the LAIP-specific combinations of antibodies were applied and 250,000 events/tube were acquired. Life-gating was not applied. Analysis of list mode files was performed with CellQuest software (Becton Dickinson). Fixed instrument settings were used for leukemic samples at both time points and normal BM samples.
LAIPs were defined by gating of a population that displayed an aberrant expression of surface or cytoplasmic antigens and by applying Boolean logic gating schemes so that the populations of interest were electronically separated for independent analysis. LAIPs were grouped into (a) cross-lineage antigen expression, (b) asynchronous antigen expression, (c) antigen overexpression, and (d) lack of antigen expression. The combination of gates obtained by this strategy was applied to the list mode data files that contained data from normal BM samples that had been collected by using the same combinations of antibodies. The frequencies of cells within the normal BM samples carrying the respective LAIPs were determined for each LAIP. The immunophenotype of AML is frequently heterogeneous, and, in many cases, several leukemic subpopulations can be detected. Therefore, it is not possible to include all leukemic cells into one LAIP in these cases, and the frequencies of LAIP-positive cells are lower in contrast to acute lymphoblastic leukemia, which generally displays homogeneous populations.
By this approach, the list mode data files obtained from BM samples at the time of relapse were evaluated for each LAIP defined at the initial diagnosis. If the initially aberrant phenotypes disappeared at the time of relapse, new aberrations by applying the same approach were identified.
The threshold for the definition of a phenotype shift at relapse was considered to be fewer than 1% of cells at relapse that were positive for the initially defined LAIPs. In accordance with our previous analyses, we calculated the logarithmic parameter: log difference = frequency LAIP-positive leukemic cells at diagnosis (relapse)/median frequency of LAIP in normal BM (16).
Cytomorphology, Cytogenetics, and Molecular Genetics
Assessment by cytomorphology and cytochemistry was based on May-Grünwald-Giemsa stain, myeloperoxidase reaction, and nonspecific esterase using α-naphtyl-acetate. All stains were performed routinely according to standard procedures (53–55). AML was diagnosed by cytomorphology and cytochemistry according to criteria defined in the French-American-British classification (50, 56, 57). Relapse was defined according to published criteria (58). Cytogenetic analyses were based on standard protocols as previously described (59). Cytogenetic data were classified according to the International System for Human Cytogenetic Nomenclature (21, 60, 61). Molecular genetic analyses were performed as described previously (62, 63). The samples were analyzed for the fusion gene transcripts (CBFB-MYH11, PML-RARA, AML1-ETO, AML1-EV1, and MLL-AF9), for MLL-PTD and FLT3-LM, and for mutations involving codons D835 and I836 of the FLT3 gene (FLT3-TKD).
Numerical values were compared with Student's t test. Pearson's chi-square test was used to test for differences in the distribution of dichotomous variables. Statistical analysis was performed with SPSS 11.0.1 (SPSS Inc., Chicago, IL, USA). All reported P values are two-tailed.
Forty patients (82%) had de novo AML and nine (18%) had secondary AML (median age 58 years; range 19–81 years). The median duration from diagnosis to relapse amounted to 9.2 months (range 3.4–30.9 months). The median percentages of leukemic blasts in BM were 80% (range 9–100%) at diagnosis and 70% (range 6–97%) at relapse. In all cases, at least one LAIP was defined at diagnosis. If more than three LAIPs were present in a single patient at diagnosis, the three most sensitive LAIPs were selected for further analysis based on the maximum log difference for comparison to normal BM.
Presence of LAIPs at Relapse
In 37 of 49 patients (76%), at least one LAIP detectable at diagnosis was present at relapse, with only a slight change in the percentage of aberrant subpopulations. In 12 patients (24%), none of the initial LAIPs was present in more than 1% of marrow cells. In seven of these 12 patients, at least one additional LAIP was acquired at relapse. However, there were insufficient numbers of cells in the remaining five patients for a complete analysis, and only staining with the triple combination selected for the identification of the respective LAIPs present at diagnosis was used. Therefore, the true number of cases in which none of the LAIPs present at diagnosis was present at relapse was seven of 44 (16%).
A total of 99 LAIPs was identified at diagnosis (median 2 LAIPs/patient). Most of the LAIPs were defined as overexpression of antigens (n = 38). Others were cross-lineage antigen expression (n = 29), lack of antigen expression (n = 24), and asynchronous antigen expression (n = 8). The median percentages of BM cells carrying the respective LAIPs were 15.90% (range 2.31–67.69%) at diagnosis and 4.43% (range 0.004–62.12%) at relapse. Overall, 30 of 99 LAIPs present at diagnosis (30.3%) were not detectable in more than 1% of BM cells at relapse.
Patterns of Changes in LAIPs
Based on the changes in LAIPs, three groups of patients were identified. In the first group, there was no change in any LAIP at relapse (Fig. 1). The most heterogeneous second group contained patients who expressed at least one of the initially detected LAIPs at relapse, whereas some LAIPs disappeared, and others were acquired at relapse (Fig. 2). In the third group, none of the patients at relapse displayed the LAIPs present at diagnosis, but additional LAIPs were observed (Fig. 3). There were no differences in the percentages of leukemic blasts in BM at relapse across these three groups. However, there were significant differences with regard to changes in findings other than the flow cytometric ones. Thus, the respective frequencies for groups 1, 2, and 3 were 11%, 40%, and 58% for morphologic changes (P = 0.007); 15%, 20%, and 25% for cytogenetic changes (not significant); and 18%, 0, and 86% for molecular genetic changes (P = 0.002).
Group 1: no change in LAIPs.
A total of 56 LAIPs in the first group of 27 patients was analyzed (Table 1). The median percentages of the aberrant leukemic subpopulation at diagnosis were 12.01% (range 4.35–80.0%) and 9.85% (range 1.09–61.99%) at relapse. A change of blast cytochemical characteristics was described in only three patients (11%), all of which were only minimal (from M1 to M2, n = 2, and from M0 to M1, n = 1). A change of the karyotype occurred in four patients (15%). In three of them, an evolution of the initially present aberrant karyotype was observed.
Table 1. Characteristics of Patients With Unaltered Immunophenotype at Relapse*
Group 2: at least one LAIP constant between diagnosis and relapse.
In the second group of 10 patients, the presence of 23 LAIPs between diagnosis and relapse was compared (Table 2). The median frequencies of the LAIPs analyzed at diagnosis and at relapse were 18.34% (range 3.82–60.49%) and 2.03% (range 0.004–34.17%), respectively. The median duration from the initial diagnosis to relapse was significantly shorter than that in the first group of patients (median 7.0 vs. 10.3 months; P = 0.025). Three of 10 patients had two LAIPs detectable, and the remaining ones disappeared at relapse. In five patients, only one of initially observed LAIPs was present at relapse. The two remaining patients initially displayed only one LAIP, which was also present at relapse, although additional LAIPs were observed at relapse.
Table 2. Characteristics of Patients With Partial Shift of Aberrant Immunophenotype
The subgroup of LAIPs in which LAIPs disappeared at relapse most frequently was the group with overexpression (six of 12). However, although 10 of the overall 23 LAIPs were present at relapse in fewer than 1% of BM cells, in four cases these LAIPs were still present in higher frequencies as compared to normal BM (log difference > 1.0). Relatively small changes in blast cytochemical characteristics occurred in four patients (from M1 to M2, n = 2; from M6 to M2, n = 1; and from M4 to M1, n = 1) and changes in karyotype in two of these four patients.
Group 3: change in LAIPs.
In 12 patients, a loss of all previously present LAIPs was observed and new LAIPs were acquired (Table 3). The median frequencies of LAIPs at diagnosis and at relapse were 18.36% (range 2.31–48.54%) and 0.01% (range 0.004–0.66%), respectively. Antigen overexpression was the most common (n = 11) and asynchronous antigen expression was the least frequent subgroup of LAIPs (n = 1) within these 20 LAIPs. Interestingly, for 20% of these LAIPs, the frequency of LAIP-positive cells was one log higher than in normal BM. The duration of CR did not differ significantly between the third and first or second groups, respectively (median 8.5 vs. 10.3 months and 8.5 vs. 7.0 months). Aside from the shift of LAIPs, changes by at least one of the other cellular characteristics were detected at relapse in this group in 11 patients (92%; P = 0.003 vs. groups 1 [33%] and 2 [40%]). Changes in blast cytochemical characteristics were observed in seven patients (58%). In particular, initial M5a morphology of blasts changed in two patients to M1 and M2 morphology, respectively, reflecting the occurrence of a secondary AML (Fig. 4). A change in karyotype occurred in four of 12 patients (33%). In contrast to the other two groups, the karyotype changes at relapse were described in two patients with an initially normal karyotype. In five of seven patients (71%) assessed for molecular alteration at diagnosis and at relapse, changes were observed (change in FLT3-LM in three patients and change in FLT3-TKD mutation in two patients). Changes detected by the different methods were not coincident within distinct cases but also occurred for one method without changes for the other methods.
Table 3. Characteristics of AML Patients With Different Immunophenotypes at Diagnosis and at Relapse
Because of increasing importance of MRD in the risk-adapted management of patients with AML, we focused our study on the incidence of LAIP changes between diagnosis and relapse. The LAIPs were established at diagnosis in each patient by applying an extensive panel of 31 monoclonal antibodies in triple combinations. The findings by flow cytometric analysis were compared with the morphologic, cytogenetic, and molecular genetic analyses of leukemic cells. Several studies have reported on the instability of immunophenotype between diagnosis and relapse in a significant number of cases (28, 31, 64–69). However, no analyses are available on the incidence of LAIP changes that also detail the development of changes in other cellular characteristics at relapse.
In the present study, 99 LAIPs in 49 patients were identified at diagnosis. In 76% of patients (n = 37), at least one LAIP remained detectable at relapse. The incidence of changes in LAIPs was strongly correlated to the incidence of morphologic, cytogenetic, and molecular genetic changes. In 12 patients (24%), none of the initially identified LAIPs could be detected in more than 1% of BM cells at relapse. To exclude that the disappearance of LAIPs occurred solely because of an insufficient sensitivity of LAIPs, comparisons of the sensitivities of each LAIP were performed across the three groups, with no changes, some changes, and significant changes in LAIPs. LAIPs were most sensitive in group 3; thus, the reason for the absence of these LAIPs at relapse was due to a change in phenotype. Most interestingly, however, the most changes in blast cytochemical characteristics, karyotype, and molecular genetics occurred in this group. Taking into account these findings, only one of these patients had a change of LAIPs in the absence of other changes. Based on this observation, it is suggested that a different leukemia developed during the antileukemic therapy, i.e., an AML resulting from clonal evolution or even a treatment-related AML. Therefore, any of the currently available methods for MRD detection would not be capable of identifying these patients. Clearly, the flow cytometric assessment of immature cells by applying CD45 right-angle gating should be included in follow-up analyses, which is still considered more sensitive than cytomorphology (70).
In 55% of patients (group 1), all of the initial LAIPs remained detectable at relapse; however, the percentage of LAIPs changed in some patients. Nevertheless, the lack of LAIP-positive populations was not complete in all the other patients (groups 2 and 3). The level of LAIP-positive cells at relapse in 40% of these patients was at least one log higher than in normal BM. Therefore, even though there was a change in immunophenotype between diagnosis and relapse, the application of MFC to monitor MRD would still show abnormal levels and thus be feasible even in these cases.
The group of LAIPs within which changes in LAIPs occurred most frequently between diagnosis and relapse was the overexpression of antigens (42%). This association was anticipated because a LAIP is defined basically on the expression intensity of one or two antigens only in some cases. It is expected that these limitations will be resolved by technical improvements. Along this line, in particular the use of five fluorescence dyes that allow the definition of more characteristics of one LAIP is considered essential. An increase in sensitivity and accuracy of MFC-based MRD monitoring will also be achieved by applying CD45 gating in the analysis of the data.
In conclusion, it is possible to detect an impeding relapse in 80% of AML patients by MFC-based monitoring of MRD. In some of these patients, there are minor changes in the immunophenotypic aberrations, the relevance of which could be decreased by improvement of technical aspects for detection of LAIPs. In a minority of patients, the occurrence of secondary AML rather than relapse is a challenge for all approaches of MRD monitoring (51).
The authors thank Karin Hecht, Rita Lapping, and Eva Goecke for excellent technical assistance. The authors greatly acknowledge all of the more than 300 physicians for their confidence in our laboratory, for sending us the patient samples, and for providing us with information on the clinical courses of patients.