Acute megakaryoblastic leukemia (FAB M7) accounts for approximately 6% of childhood acute myeloid leukemia (AML) cases (). Although this subtype is rare and has a dismal outcome in pediatric patients without constitutional chromosomal abnormalities, it is the most frequent one in patients with Down syndrome (DS) (). Patients with DS not only show an increased incidence of leukemia compared to the general pediatric population but also about 5–10% of all infants with DS develop a transient myeloproliferative disorder (TMD) characterized by clonal proliferation of typically megakaryoblastic or erythroblastic myeloid blasts ([3-6]). In most patients this disease is self-limiting and disappears within the first few months of life without treatment. However, between 13 and 33% of patients with TMD develop AML in the first 4 years which has a rather good prognosis when treated properly ([5, 7, 8]).
Although most AML FAB subtypes can be diagnosed reliably with morphology alone, the classification of AML FAB subtypes M0 as well as FAB M7 relies on immunophenotypic assessment ([9, 10]). According to the European Group for the Immunological Characterization of Leukemias (EGIL) classification, CD41 and CD61 are considered as markers of the megakaryocytic lineage (). CD41 is expressed throughout all stages of maturation of platelets and has therefore been described as a specific marker for AML-M7 and TMD ([12-14]). However, adherence of platelets to leukocytes including monocytes and blast cells can lead to false positive results for this marker ([15, 16]). These technical difficulties can lead to erroneous diagnosis of AML M7.
We reported earlier that 33% of pediatric B cell precursor acute lymphoblastic leukemia lacked expression of CD11a, an aberration exploitable for minimal residual disease (MRD) detection (). We were therefore interested in the expression of this marker also on blast cells of AML patients.
Human lymphocyte function-associated antigen 1 (LFA-1, CD11a) functions both as a key adhesion receptor in immune and inflammatory processes as well as a signal-transducing molecule ([18, 19]). CD11a is expressed on normal leukocytes including most normal bone marrow (BM) CD34+ cells. The intensity of expression increases with normal leukocyte differentiation while this marker is lost in an early step of erythropoesis ([17, 19, 20]). Hence, we reasoned that CD11a could be differentially expressed also in AML subtypes. One study reported CD11a to be widely expressed on the majority of AMLs (). However, there are no data available on a consistently deficient expression in a specific subtype of AML.
In the following we show, that deficiency of CD11a is a sensitive and specific finding in pediatric AML-M7 as well as TMD in contrary to AML of other subtypes. We propose that assessment of CD11a deficiency can be used as a new diagnostic and follow-up marker for these entities.
- Top of page
- MATERIAL AND METHODS
- LITERATURE CITED
- Supporting Information
For patient characteristics see Supporting Information Tables S1 (study group) and S2 (control group). Our cohort mainly reflects the distribution of FAB subtypes of the whole AML-BFM 89/2004 cohort. Only the percentage of TMD and M2 patients was statistically higher in our group than in the whole cohort (P = 0.007 and 0.04), for the other subtypes, percentages were not statistically different.
Median age at diagnosis of TMD was 0.02 years or 7.3 days (mean 0.04, range 0.00–0.12) comparable to earlier data (), which was significantly lower than for patients with DS-AML (1.73, mean 1.97, range 0.02–4.87, P < 0.001) and Non-DS M7/M6 (1.37, mean 1.43, range 0.91–2.60, P < 0.001). Klusmann et al. found a comparable age at diagnosis of DS-AML with 1.5 years (). Only one child diagnosed with DS-AML was older than 4 years at diagnosis which is consistent with earlier reports ([2, 3]). None of the DS patients carried the typical favorable AML aberrations t(8;21), inv(16) or t(15;17) similar to previous data ().
We investigated and semi-quantified CD11a expression in the leukemic samples. All eight patients with Non-DS M7, the sole patient with M6 and all 27 patients with DS showed deficiency of the antigen (Fig. 1). The median MFI of CD11a on blast cells was very low in TMD (4; range 3–9), DS-AML (4.5; range 3–15) and Non-DS M7/6 (6; range 4–9) while it was significantly higher with a median value of 585 in the control group of AML M0-M5 (range 8-6066, P < 0.001). The sensitivity of deficient CD11a expression for the diagnosis of DS-TMD/AML or Non-DS M7/M6 was 100% (36/36) and the specificity 95% (CD11a negative in 3/55 Non-DS M0-M5 patients). Again, this difference in expression was highly significant (P < 0.001). When evaluating the Non-DS M0-M5 group for FAB-subtype associated CD11a expression peculiarities (Fig. 2), we found that in particular monocytic leukemias (M4/M5) showed very high MFI values (median 1,258, range 168–6,066), M1/M2 were intermediate (MFI median 348, range 8–1,490), and all five cases with AML-M3 were rather low in CD11a expression (median MFI 24, range 17–39, two cases classified negative). Excluding these latter cases from the control group, specificity of deficient CD11a expression was 98% (CD11a negative only 1/50 Non-DS M0/1/2/4/5 cases). Analysis of positive/negative assessment based on a percent-based threshold (i.e., 20%), a very similar pattern of expression was found: 35/36 samples of the group with DS and non-DS M6/M7 could be classified as being negative for CD11a (the sole positive sample showed 22% positive blasts and a MFI of 15), only 2/5 Non-DS M3 cases were positive, and of the remaining Non-DS M0-M5 48/50 were positive. The two discrepant Non-DS M0-M5 cases were both marginally positive by MFI (24 and 27), but negative by percent-threshold.
Figure 1. CD11a expression in pediatric AML samples. (A) All patients with Down syndrome (n = 27) and Non-Down syndrome AML M7/6 (n = 9) showed deficient CD11a expression. Deficiency of CD11a expression was significantly less frequent in Non-Down syndrome M0–5 patients (n = 55, P < 0.001). (B) CD11a expression of FAB subgroups: All TMD and M7 patients exhibit CD11a deficiency. Note that DS-AML cases FAB M0 are also CD11a-deficient.
Download figure to PowerPoint
Figure 2. Levels of CD11a expression represented by MFI values of normal hamatopoietic cells and leukemic cells of different FAB subtypes.
Download figure to PowerPoint
We were also interested in assessing the physiologic CD11a expression characteristics of normal hematopoietic counterpart cells as opposed to leukemic expressions (see also Fig. 2). Normal erythroblast cells were totally devoid of CD11a expression (median MFI 3; range 2–5), whereas normal neutrophils in the study group had an intermediate median MFI of 367 (range 183–1,253) similar to neutrophils from healthy donor samples (median MFI 431; range 224–643). The median in-sample ratio between the MFI of M7/TMD blast cells and neutrophils (available in 31/36 patients) was 1:92 (range 16–271). Normal monocytes exhibited a very high CD11a expression (median MFI 2822, range 2,397–3,824). Normal myeloid CD34+ precursors in BMs of non-leukemic patients exhibited a rather high CD11a expression (median MFI 744; range 572–1,064), whereas normal B-lymphoid precursors (CD34+) were very low (37; 28–50) as described earlier ().
With regard to the usage of CD11a as a marker for MRD detection, we investigated CD11a expression on regenerating BM cells of patients with nonmyeloid acute leukemia or lymphoma. Among n = 9 BM samples the median percentage of CD11a-deficient CD34+ cells was only 5.2% (2 SD: 5.4) and the median percentage of CD34+CD11a-deficient cells of all nucleated BM cells was 0.052% (2 SD; 0.074) (Table 1). These CD11a-deficient normal cells were B-lymphoid precursors throughout.
Table 1. CD11a Expression on CD34+ Regenerating Bone Marrow Cells
|BM sample||CD34+/all NC (%)||CD34+ CD11a−/ all NC (%)|
As an example of the usefulness of CD11a for diagnostic work-up and MRD follow-up in patients with AML M7 or TMD, one patient was investigated in detail (Fig. 3). The patient was diagnosed with marked thrombocytopenia at 8 months of age and had 0.5% CD34+CD7+CD56+/−CD41− blasts in PB which were CD11a-negative (Fig. 3A). He was followed without intervention until developing more pronounced signs of GATA1-mutated (exon 2, 335 dup GT) AML M7 at the age of 13 months with 15% blasts in BM (Fig. 3B). Blast cells were CD11a-deficient and co-expressed CD7, CD56, and CD41. CD41 was falsely positive also on monocytes (blue) and few lymphocytes (green), which is a well-known technical draw-back of this marker due to platelet adherence to normal cells. After induction chemotherapy with two courses the patient was in remission with very active hematopoietic regeneration. Nevertheless, a doubtful CD34+ cell population was seen with co-expression of CD56 and CD7 (Fig. 3C). Strong expression of CD11a, however, suggested a nonleukemic derivation which was proven by a sorting experiment and negative result in subsequent GATA1 mutational analysis.
Figure 3. Immunophenotypic data of a patient with DS-AML. (A) Patient at diagnosis of TMD at 8 months with 0.5% CD34+CD7+CD56+/−CD41− blasts in PB which were CD11a-negative (red subset). (B) Development of AML-M7 at 13 months, 15% blasts in BM (CD7+, CD56+, CD41+, CD11a-, red), CD41 is falsely positive on monocytes (blue) and few lymphocytes (green). (C) Hematopoietic regeneration after induction chemotherapy. Coexpression of CD56 and CD7 on CD34+ cells, which were CD11a-negativity. GATA1 analysis after sorting proved non-leukaemic derivation of these cells.
Download figure to PowerPoint
Next, we investigated further marker characteristics of the AML-M7/M6/TMD cases (Table 2). Although per definition, all cases with DS-M7 and Non-DS M7 were positive for CD41 and/or CD61, only 62% of DS-TMD patients (8/13) were considered positive in CD41 (7/13) and/or CD61 (2/6) above the cut-off value of 20%. The mean percentage of CD41 or CD61 positive blasts in the cases regarded as negative per definition was 12.7 and 13.5%, respectively. Two of four DS-M0 cases also had 15 and 16% CD41 positive blasts (negative per definition). HLA-DR deficiency was also not a consistent marker of the study cohort: we found it expressed in 26% of cases with AML-M7/M6/TMD and in the majority of the other AML cases (48/55, 87%). The difference in frequency, however, was statistically significant (P < 0.001). Notably, the same seven patients in the M0–M5 group who were HLA-DR negative were CD11a negative, too, including the five M3 cases.
Table 2. Surface Marker Expression of the Study and Control Group
| ||CD11a||CD41||CD61||CD34||CD117||MPO||CD13||CD33||CD56||CD7||HLA DR|
|DS-M0 (n = 6), n (%)||0/6 (0)||0/4 (0)||0/1 (0)||5/6 (83)||6/6 (100)||0/4 (0)||6/6 (100)||6/6 (100)||3/4 (75)||5/6 (83)||0/6 (0)|
|DS-M7 (n = 6), n (%)||0/6 (0)||5/6 (83)||2/2 (100)||3/6 (50)||5/6 (83)||0/5 (0)||5/6 (83)||6/6 (100)||3/4 (75)||6/6 (100)||3/6 (50)|
|DS-AML (n = 12), n (%)||0/12 (0)||5/10 (50)||2/3 (67)||8/12 (67)||11/12 (92)||0/9 (0)||11/12 (92)||12/12 (100)||6/8 (75)||11/12 (92)||3/12 (25)|
|TMD (n = 15), n (%)||0/15 (0)||7/13 (54)||2/6 (33)||13/15 (87)||10/10 (100)||0/9 (0)||3/14 (21)||14/15 (93)||3/7 (43)||13/14 (93)||4/14 (29)|
|DS total (n = 27), n (%)||0/27 (0)||12/23 (52)||4/9 (44)||21/27 (78)||21/22 (95)||0/18 (0)||14/26 (54)||26/27 (96)||9/15 (60)||24/26 (92)||7/26 (27)|
|Non-DS M7/M6 (n = 9), n (%)||0/9 (0)||7/8 (88)||6/6 (100)||2/9 (22)||4/8 (50)||0/8 (0)||3/8 (38)||7/9 (78)||2/8 (25)||3/9 (33)||2/8 (25)|
|Total TMD/M7/M6 (n = 36), n (%)||0/36 (0)||19/31 (61)||10/15 (67)||23/36 (64)||25/39 (83)||0/26 (0)||17/34 (50)||33/36 (92)||11/23 (48)||27/35 (77)||9/34 (26)|
|Non-DS M0-M5 (n = 55), n (%)||48/55 (87)||0/45 (0)||0/11 (0)||36/55 (65)||30/54 (56)||40/52 (77)||35/55 (64)||47/55 (85)||6/12 (50)||22/55 (40)||48/55 (87)|
CD117 was expressed on almost all patients with DS (21/22, 95%), while in Non-DS M7/6 and M0-M5 cases, this marker was expressed in a smaller percentage (4/8, 50% and 30/54, 56%). Comparably, most cases of DS were positive for CD7 (24/26, 92%), while only 33% (3/9) of cases with Non-DS M7/M6 and 40% (22/55) of cases with Non-DS M0-M5 were positive. Most blast cells of DS-AML patients expressed CD56 (6/8, 75%). For TMD patients (3/7, 43%), Non-DS M7/M6 (2/8, 25%) and Non-DS M0-M5 (6/12, 50%), these numbers were lower.
- Top of page
- MATERIAL AND METHODS
- LITERATURE CITED
- Supporting Information
Flow cytometric evaluation of surface and cytoplasmatic markers is an indispensable method to characterize subgroups of acute leukemias and may also be a complementary prognostic tool ([11, 26, 27]). A marker expressed on more than 20% of blasts cells is considered positive, with the exception of highly specific markers like MPO, cytoplasmatic CD3 and CD79a with a cut-off of 10% (). However, interpretation of percentages just below or above the cut-off line is disputable. In addition, technical difficulties lead to results that are not always definite. Although the FAB classification of M1 to M6 is still mainly based on morphology and cytochemistry, flow cytometry is needed for the accurate classification of M0 and M7 (). An expression of either CD41, CD42 or CD61 is required for the diagnosis of AML-M7 ([9, 10]). CD41 has also been described as a marker for TMD ([7, 13, 29]). Technical problems, including adherence of platelets to leukocytes including monocytes and blast cells can lead to false positive results for these markers (). In our experience, this phenomenon leads to inconclusive results in a considerable number of cases (Fig. 3B). Karandikar et al. examined the immunophenotype of 18 patients with TMD and DS-AML and found inconclusive results for the expression of CD41 or CD61 in 4/18 cases (22%) due to background staining of neutrophil leukocytes and monocytes (). Langebrake et al described expression of CD41 or CD61 in only 57% of TMD and 65% of DS-AML patients, Klusmann et al. in 62% of TMD patients ([13, 25]). These technical imprecisions and the lack of sensitivity for known TMD and M7 markers underlie the need for new markers in the diagnosis of these entities.
CD11a plays a major role in mediating and stabilizing leukocyte adhesion (). In this study, which included a large (46%) and representative cohort of all pediatric AML cases diagnosed in Austria since 1998, we found a consistent deficiency of CD11a in all patients with DS-TMD, DS-AML, as well as in Non-DS AML M7 and the sole case of AML M6. In comparison with the CD11a expression levels of other AML subtypes, the sensitivity of this pattern was very high for M6/M7/TMD with 100%. Specificity was high with 95% as only three of 55 patients with Non-DS M0–M5 showed deficient CD11a MFI values, including two patients with M3 subtype investigated in this study. Zangrando et al. already described a low CD11a expression in M3 (). However, as M3 leukemias are mainly identified by specific morphologic, cytochemical and genetic characteristics, their lack of CD11a expression will not lead to difficulties in distinguishing M3 cases from cases with TMD or M7. Hence, the specificity of CD11a was 98% when excluding M3 cases from assessment. To apply these findings properly, it has to be kept in mind that MFI values are relative and depend on the flow cytometer being used. For digital flow cytometers, higher MFI values are to be expected. However, the relation of expression levels among different cell subtypes should be comparable.
In contrast to a clear deficiency of CD11a, results for expression of CD41 and CD61 were inconsistent in our study group. Although per definition all cases with DS-M7 and Non-DS M7 were positive for CD41 and/or CD61, only 62% of DS-TMD patients showed expression of CD41 and/or CD61 above the cut-off value. However, in all of these latter patients as well as in two of four DS-M0 patients, these markers were expressed in a percentage that just missed the cut-off criterion of 20% (the two other DS-M0 cases were only assessed for one of CD41/CD61 and found negative for the antigen). Hence, despite morphology conspicuous of M7, these cases were diagnosed with M0 instead of M7 just due to the lack of immunological criteria. This partial lack and inconclusivity of immunological data may contribute to the high proportion of M0 cases diagnosed in our DS-AML group (6/12 cases) while Creutzig et al. observed M7 to be the most frequent subtype of DS-AML (86%) in their much larger cohort ().
Langebrake et al. investigated the expression profile of a large cohort of patients with DS-TMD and DS-AML and described a characteristic expression pattern of CD33+/CD117+/CD34+/CD7+/CD56+ (). According to our results, none of these markers can be used to unambiguously distinguish such cases from AML M0–M5 subtypes as they were all expressed on a considerable number of these subtypes.
As a more characteristic feature of DS and Non-DS M6/7 we found absence of HLA-DR in 74% of cases (25/34) while this marker was positive on blasts of the majority of the other AML cases (48/55, 87%, P < 0.001). Notably, all CD11a negative patients from our control group were also negative for HLA-DR, including two AML-M3 cases. HLA-DR deficiency in M3 has been described before ([28, 31]). In a study by Karandikar et al., most of the DS-AML cases were HLA-DR negative (6/9, 67%), however a much smaller number of TMD cases (2/9, 22%). Langebrake et al. showed a significantly less frequent expression of HLA-DR in DS cases and also their Non-DS M7 cases were mainly HLA-DR negative (). Although absence of HLA-DR is not as sensitive and specific as deficiency of CD11a, this characteristic feature may be used for the diagnosis of DS-TMD/AML and Non-DS M7 in conjunction with CD11a.
Differentiation of hematopoietic cells in the different lineages is accompanied by changes in the antigen profile including CD11a (). In accordance with these reports, CD34+ myeloid precursors, normal neutrophil granulocytes and cases of AML M1/M2 showed intermediate to high CD11a expression while this marker was low on CD34+ B lymphoid precursor cells (Fig. 2). Normal PB monocytes and also monocytic AML cases (M4/M5) in our study exhibited as a group the highest CD11a expression of all hematopoietic cells investigated, while in contrast CD11a expression was absent on erythroblasts.
There has been evidence that erythroid and megakaryocytic differentiation pathways are closely related: Cases of AML with overlapping phenotype of M6 and M7 have been reported ([33, 34]). In addition, there is evidence that blasts in M7 and TMD may correspond to progenitors of both the erythroid and megakaryocytic lineage (). In light of close relations of these two lineages, we included the sole case of AML-M6 in our study cohort. However, due to the paucity of M6 cases in our pediatric cohort, we can only speculate that this subtype might consistently be CD11a deficient similar to M7.
With respect to differences between the CD11a-deficient entities, there is clear evidence that DS AML-M7 is distinct from AML-M7 in Non-DS patients in many aspects including cytogenetic abnormalities, response to chemotherapy and course of the disease (). At least with respect to immunophenotype, our study now shows that Non-DS and DS-AML/TMD share certain antigen expression similarities, suggesting that these entities might still arise from the same precursor cell with the potential to differentiate into the megakaryocytic and the erythroid lineages.
A remaining question is whether TMD blasts can be distinguished from “real” AML by any means in patients with DS. It was proposed that AML in DS patients arises from the leukemic cells of TMD following a second hit. Morphologically, blasts in both diseases are indistinguishable and exhibit features of megakaryoblasts (). They also share the same cytogenetic changes, and the same GATA1 mutations were found in both TMD blasts and AML blasts ([5, 25]). In our cohort, CD11a expression was missing uniformly in all DS-TMD and DS-AML patients. In one patient with paired samples from TMD diagnosed on the second day of life and from AML breaking out clinically at the age of 2.1 years, the immunophenotypic profiles were very similar. Overall, we were not able to determine a specific phenotype which would allow distinguishing TMD from AML, apart from a significantly more frequent expression of CD13 on DS-AML blasts than on TMD blasts (11/12 vs. 3/14, P > 0.01).
Detection of MRD in AML patients during treatment follow-up correlates with outcome. Hence, there is a clear trend to use MRD end-points for better risk stratification and treatment selection in pediatric and adult AML ([6, 32, 33, 35, 36]). Flow cytometry is the technical approach which has the broadest applicability and least need for a per patient test individualization among all currently available methods. It has also a short turn-around time so that results are available within one day. Nevertheless, there is a need for markers or antigenic constellations which are specific and sensitive enough to allow discriminating MRD from normal precursor cells at sufficient resolution. Our study now shows that deficiency of CD11a on blasts of AML-M7/M6/TMD is such a highly sensitive and specific marker for these disease entities. The percentages of normal CD34+ cells deficient of CD11a among all nucleated cells were low (median 0.052%) even in post-chemotherapy regenerating BMs. As these normal precursors were B-lymphoid precursors throughout, inclusion of markers like CD10 or CD19 into AML-MRD panels using CD11a should allow increasing its sensitivity even further. We propose that CD11a should be central to attempts towards assessing MRD in AML M7 and typical DS-AML, respectively. As shown in Figure 3, this marker should also be very useful for diagnosing and quantitatively following the evolution of TMD in DS infants. Importantly, a CD11a-based approach will allow detecting correctly the entire blast population. As also exemplified in Figure 3, this is usually not possible with CD41 and similar markers because these antigens are frequently positive only on subpopulations of the pathological cell clones.
In summary, we propose that deficiency in CD11a expression should be added to the diagnostic criteria of AML-M7, classical DS-AML and TMD. Our preliminary data also strongly suggest application as MRD marker of these disease entities, however the exact role as MRD marker has to be evaluated in further studies.