Immunological classification of acute myeloblastic leukemias - Nature

Leukemia (2003) 17, 515–527 & 2003 Nature Publishing Group All rights reserved 0887-6924/03 $25.00 www.nature.com/leu

Immunological classification of acute myeloblastic leukemias: relevance to patient outcome RO Casasnovas1, FK Slimane1, R Garand2, GC Faure3, L Campos4, V Deneys5, M Bernier6, A Falkenrodt7, G Lecalvez8, M Maynadie´1 and MC Be´ne´3 1 He´matologie Clinique, CHU Dijon, France; 2Laboratoire d’he´matologie, CHU Nantes, France; 3Laboratoire d’immunologie, CHU Nancy, France; 4Laboratoire d’he´matologie, CHU St-Etienne, France; 5Laboratoire d’he´matologie, Universite´ Catholique de Louvain, Belgium; 6Laboratoire d’he´matologie, Institut J. Bordet Bruxelles, Belgium; 7Laboratoire d’he´matologie, CHU Strasbourg, France; 8Laboratoire d’he´matologie, CHU Brest, France

Immunophenotyping is a major tool to assign acute leukemia blast cells to the myeloid lineage. However, because of the large heterogeneity of myeloid-related lineages, no clinically relevant immunological classification of acute myeloblastic leukemia (AML) has been devised so far. To attempt at formulating such a classification, we analyzed the pattern of expression of selected antigens, on blast cells collected at AML diagnosis. Patients were eligible if they had a first diagnosis of de novo AML and a sufficient number of blast cells for proper immunophenotyping. The relative expression of CD7, CD13, CD14, CD15, CD33, CD34, CD35, CD36, CD65, CD117, and HLADR were analyzed by cytometry in a test series of 176 consecutive AML cases. Statistical tools of clusterization allowed to remove antigens with overlapping distribution, leading us to propose an AML classification that was validated in a second AML cohort of 733 patients. We identified five AML subsets (MA to ME) based on the expression of seven antigens within four groups (CD13/CD33/CD117, CD7, CD35/CD36, CD15).-MA and MB-AML have exclusively myeloid features with seldom extramedullary disease and rare expression of lymphoid antigens. No cases of acute promyelocytic leukemia (APL) were observed within MB AML. MC AML have either myeloid or erythroblastic features. MD AML have more frequently high WBC counts than other subsets, which were related to the expression of CD35/CD36 and CD14 and to monoblastic differentiation. ME AML lack CD13, CD33, and CD117 but display signs of terminal myeloid differentiation. Specific independent prognostic factors were related to poor overall survival in each immunological subset: CD34+ (Po3 104) in MA AML, CD7+ in MB AML, non-APL cases (Po0.03) in MC AML, CD34+ (Po0.002) and CD14+ (Po0.03) in MD AML, CD14+ in ME AML (Po0.01). The inclusion of seven key markers in the immunophenotyping of AML allows a stratification into clinically relevant subsets with individual prognostic factors, which should be considered to define highrisk AML populations. Leukemia (2003) 17, 515–527. doi:10.1038/sj.leu.2402821 Keywords: AML; immunophenotyping; classification; prognosis

Introduction The immunological analysis of a limited panel of surface or intra cellular antigens allows to assign most cases of acute leukemia to a specific lineage. Immunophenotyping of blast cells has thus become an elective and useful tool to characterize the myeloid or lymphoid origin of blast cell populations. The use of more extensive panels further allows us to identify either undifferentiated acute leukemias(1,2) or acute leukemia entities expres-

sing simultaneously several lineage antigens, defining multiphenotypic acute leukemia.(1,3) In lymphoid lineage-derived acute leukemia, extended immunophenotyping, moreover, allows to stratify distinct clinical and biological subsets, based on the hierarchy of antigen expression during the maturation of the normal cell counterparts of blast cells.(4,5) In acute non-lymphoblastic leukemia, several antigens (myeloperoxidase, CD13, CD33, CD117) have also been defined as hallmarks of myeloid differentiation and are specifically useful for the identification of morphologically undifferentiated AML cases or the detection of blasts of either erythroid (Glycophorin, CD36) or megakaryoblastic lineage (CD41, CD42, CD61). However, although the coexpression of numerous antigens provides insights in to the myeloid lineage commitment of blast cells at various levels of maturation, because of the wide heterogeneity of myeloidrelated lineages and the numerous stages of differentiation in each of them, no relevant immunological classification of AML has been devised so far. Here we report how we were able to define an immunological classification of AML, based on the pattern of expression of selected antigens, which does not overlap with the FAB classification and defines specific clinical and outcome patterns.

Patients and methods The study was initiated in 1993 by the Groupe d’Etude Immunologique des Leuce´mies (GEIL), a multicenter group that collects data from over 40 hospitals in France and Belgium. Ten teams participated in this study. Patients eligible for the study had to meet the following criteria: established first diagnosis of de novo AML, excluding therapy-related AML and AML secondary to a previous diagnosis of myelodysplastic or myeloproliferative disease. The collection of a sufficient number of blast cells for proper immunophenotyping was also mandatory, the latter allowing lineage assignment and exclusion of BAL, and also the analysis of each of the following antigens: CD7, CD13, CD15, CD33, CD35, CD36, and CD117. FAB, clinical, and outcome data were also collected for each patient. The end point of the present analysis for patient outcome was January 1st, 2001.

Test and training series Correspondence: R-O Casasnovas, He´matologie Clinique, CHU Le Bocage, Bd de Lattre de Tassigny, 21034 Dijon Cedex, France; Fax: 33 3 80 29 36 05 This work was supported by a grant from the CHU of Dijon and by the GOELAMS Received 28 May 2002; accepted 16 October 2002

The first 176 consecutively recruited patients eligible were analyzed to establish the immunological classification of AML and were considered as a test series. In a second part of the study, this classification was applied to an independent population of 733 AML patients, defined as the training series.

Immunological classification of acute myeloblastic leukemias RO Casasnovas et al

516

Most patients received intravenous standard induction treatment including a combination of anthracyclin (daunorubicine 45– 60 mg/m2/day for 3 days or idarubicine 8 mg/m2/day for 5 days) and cytosine arabinoside(100–200 mg/m2/day for 7 days) adapted according to age. Patients who achieved a complete remission (CR) received a consolidation treatment including either an allograft transplantation if they were less than 45 years old and had an HLAmatched sibling donor, or two courses of chemotherapy including the same induction anthracyclin if they had no sibling donor or were older than 45 years of age. Patients older than 70 years received, after the first course of consolidation, a maintenance therapy during 24 months. Patients with acute promyelocytic leukemia (APL) received, during induction treatment, oral all-trans-retinoic acid (ATRA) 45 mg/m2/day until they reached CR. The consolidation treatment was followed by an oral maintenance therapy during 2 years consisting of methotrexate 15 mg/m2/week associated to 6mercaptopurine 90 mg/m2/day and ATRA 45 mg/m2/day 15 days every 3 months. Some patients, because of their advanced age or their bad performance status, received only supportive care.

Cluster analysis was performed to determine which antigens among CD7, CD13, CD14, CD15, CD33, CD34, CD35, CD36, CD65, CD117, and HLA-DR best discriminated between the immunophenotypic groups of AML patients observed. The distance measured between each antigen was computed using the percentage of disagreement.(7) The linkage rule used the weighted pair group average in which the distances between clusters were calculated as the average distance between all pairs of antigens in two different clusters, weighted by the size of respective clusters.(8) The hierarchical tree of clusters was plotted using a standardized scale (distance of linkage/maximal distance 100). The closest antigens, exhibiting a distance linkage less than 70%, were merged or removed according to their myeloid lineage weight to produce a final classification with four clusters. Overall survival (OS) was calculated from the date of AML diagnosis to the date of death, the observations of patients alive at the last visit being censored. Disease-free survival (DFS) was measured from the date of CR achievement to the date of relapse or death of any cause. The observation of patients free of disease and alive were censored. OS and DFS were estimated by the Kaplan-Meier method and compared by the logrank test. Analysis of the relative value of the prognostic factors for treatment outcome were based on Cox’s proportional hazards regression models for DFS and OS. The median follow-up duration for censored patients was 3.6 years in the test series and 2.1 years in the training series. The statistical analysis was performed using the Statistica software, version 6.0 (Tulsa, OK, USA).

Immunological phenotyping

Results

The immunophenotype was performed as previously described(6) on bone marrow blast cells collected at AML diagnosis. Flow cytometry was used in all centers on blast cells gated on their light scatter characteristics, using directly conjugated monoclonal antibodies for the following antigens: CD7, CD13, CD15, CD33, CD34, CD35, CD36, CD65, and CD117. Monoclonal antibodies against CD34 (HPCA2), CD35 (J3D3), CD36 (FA6152), CD65 (VIM2), and CD117 (95C3) were aliquoted and distributed by the GEIL to each of the participating laboratories. Monoclonal antibodies against CD7, CD15, CD13, and CD33 were those routinely used by each participating laboratory. Additional antigens were also analyzed when sufficient blast cells were collected, identifying lymphoid (CD2, CD4, CD5, CD10, CD19, CD20, CD24), or myeloid (CD11b, CD14, CD16, CD18, CD41, CD42, glycophorin A) antigens or with no lineage specificity (HLA DR, TdT). Negative controls included omission and/or substitution of the monoclonal antibody by a mouse isotype matched nonrelevant immunoglobulin. A case was considered as positive when more than 20% of blast cells reacted with the monoclonal antibody, except for TdT (10% cutoff) according to the EGIL criteria.(1) The interlaboratory reproducibility of the group, tested on five different coded samples distributed to all laboratories reached 94%.

Characteristics of patients in the test series and development of the classification

The final description of the immunological subgroups identified and analysis of the response to therapy were performed on the whole series of 909 patients.

Treatment

Statistical analysis The relation between variables were analyzed by the Fisher exact test or Pearson’s w2 test for categorical variables and by the Mann–Witney U-test or the Student’s T-test for continuous variables, depending on the sample size. Leukemia

The characteristics of the first series of 176 patients are detailed in Table 1. The classification was established by a cluster analysis procedure according to the differential expression by the blast cells of myeloid or undifferentiated lineage antigens, respecting whenever possible, the known features of physiological myeloid differentiation from immature to mature myeloid lineage. The antigens tested were thus chosen according to their diagnosis value among pan-myeloid markers such as CD13, CD33, CD117 or according to their value at identifying myeloid differentiation subsets, such as CD15 and CD65 for granulocytic differentiation, CD14, CD35, and CD36 for the monocytic lineage and CD36 for the erythroid lineage. Three nonlineage antigens, CD7, CD34, and HLA-DR were also tested. The distance between the distribution of each pair or group of antigens was assessed and the classification was progressively simplified by a forward stepwise elimination of overlapping antigens (Figure 1). The pan-myeloid antigens CD13, CD33, and CD117 had a closely similar distribution and a wide expression since more than 95% of tested AML were positive for at least one of them. Because of their specific myeloid lineage diagnosis value,(1,9) they were finally analyzed simultaneously. The linkage of CD35 and CD36 was strong enough to amalgamate them in the same cluster, and CD14 was eliminated because of its too narrow expression. Among the granulocytic differentiation antigens studied, CD15 (86 positive cases) was preferred to CD65 (62 positive cases) because of its broader expression. Among nonlineage antigens, CD7 had the widest distance linkage with respect to myeloid antigens, compared to CD34 and HLA-DR. The definitive classification was finally based on


517 Table 1

Characteristics of patients in the test and training series Test series N

Age

Median (range)

Training series %

57 (2–87)

n

%

56 (2–88)

P NS

p60 >60

94 82

Sex

M/F

1.23

FAB

M0 M1 M2 M3 M4 M5 M6 M7 Unclassified

WBC count

Median (range)

15 (0.7–274)

17 (0.5–536)

NS

Hb (g/dl)

Median (range)

9.2 (3.8–14.3)

9.2 (2.1–15.7)

NS

Platelet count

Median (range)

Extramedullary disease

Yes No

42 89

32 68

165 359

31 69

NS NS

CD2 CD3 CD5 CD7

26 5 11 52

15 3 7 29

115 15 31 169

16 3 5 23

NS NS NS NS

CD10 CD19 CD20 CD22 CD24

6 20 8 6 16

3 11 5 4 10

34 71 28 20 57

5 9 4 5 8

NS NS NS NS NS

CD13 CD14 CD15 CD16 CD33 CD35 CD36 CD65 CD117

138 27 86 10 138 62 78 63 134

78 16 49 7 78 35 44 39 75

513 148 384 81 551 212 339 287 224

76 20 52 11 75 29 46 39 74

NS NS NS NS NS NS NS NS NS

CD34 HLA-DR

87 115

47 71

320 375

46 63

NS NS

54 46

412 321

56 44

1.13

16 47 42 16 20 20 9 0 6

9 27 24 9 11 11 5 0 3

59 (6–324)

44 193 173 56 110 112 15 8 22

NS NS NS

6 26 24 8 15 15 2 1 3

59 (3–430)

NS

Immunophenotype (% of AML cases) T lineage Ag

B lineage Ag

Myeloid lineage Ag

Nonlineage Ag

the expression of 7 antigens clustered into four categories identifying five subsets of AML, namely MA, MB, MC, MD, and ME (Figures 1 and 2). MA is defined by the sole positivity of at least one of the panmyeloid antigens (CD13, CD33, CD117). MB is related to the additional positivity of CD7. MC is associated to the coexpression of pan-myeloid antigens and either CD35 or CD36 on blast cells whatever the CD7 expression. MD is characterized

by the additional expression of CD15. Finally, ME includes AML, which express none of the pan-myeloid antigens.

Characteristics of the AML classification subsets The clinical, cytologic, and immunophenotypic features of the five AML immunological classification subsets within the test Leukemia


518

a CD7 CD14 CD35 CD36 CD15 CD65 CD13 CD117 CD33 CD34 DR 40

50

60

70

80

90

100

110

Linkage distance - (Dlink/Dmax)*100

b CD7 CD15 CD35 CD36 CD13 CD117 CD33

40

50

60

70

80

90

100

110

Linkage distance (Dlink/Dmax)*100

Figure 1

Distance linkage between 11 (a) and 7 (b) antigens expressed in 176 AML cases (test series).

CD13 or CD33 or CD117

CD7

CD35 or CD36

CD15

n

Total %

+

-

-

-

33

19

146

20

179

20

MB

+

+

-

-

10

6

42

6

52

6

MC

+

-+/ +/-

+

-

44

25

136

19

180

20

MD

+

-+/ +/-

-+/ +/-

+

84

48

370

50

454

50

ME

-

+/-+/

+/-

-+/ +/-

5

3

39

5

44

5

176

733

909

Immunological classification of AML.

and the training series were compared in order to assess the reproducibility of the classification. The characteristics of the whole AML population regarding clinical and hematological parameters were very similar in the training and test series (Table 1). The frequency of antigens expression on AML blast cells of the training and the test series Leukemia

Training series n %

MA

Total

Figure 2

Test series n %

was very close as well, and the clustering method was applied to the training series population. The segregation of the various antigens was similar to those observed in the test series (Figure 3). The main differences appeared in the linkage distances observed either among pan-myeloid antigens or between CD15 and the CD35/CD36 cluster, which were shorter


519 CD7 CD36 CD35 CD15 CD13 CD117 CD33

40

50

60

70

80

90

100

110

Distance linkage - (Dlink/Dmax)*100

Figure 3

Distance linkage between the seven antigens of the AML classification in the 733 cases of the training series.

in the test than in the training series, thus strengthening and validating the choice of clusters used for the classification. The main characteristics of MA to ME AML subsets were first established on the test series and confirmed on the larger training series. Since we did not observe any difference between the features of the immunological subsets analyzed of both series, we will report the AML subsets characteristics on the overall population of 909 patients (Table 2). Hematological parameters are similar in the different subsets except for WBC, which are significantly higher in MB and MD AML patients than in patients belonging to other AML subgroups (Po0.004). Conversely, both MA and MB AML have significantly less frequent extramedullary disease (Po0.05). MA subset: The blasts of most of MA AML patients have a myeloid morphology (72%) and were predominantly classified as M1 in the FAB classification. With regard to differentiation antigens not involved in the immunological classification, MA AML have a significantly lower frequency expression of the myeloid antigens CD14 (Po105), CD16 (P ¼ 0.0002) and CD65 (Po105), also of the nonlineage antigens HLA-DR (Po105) and CD34 (Po0.05), than the other AML subsets (Table 2). Lymphoid lineage antigens were also seldom expressed in this subset. MB subset: MB AML represent the smallest group of AML expressing pan-myeloid antigens. Despite CD7 coexpression, MB AML do not usually express other T lymphoid lineage antigens such as CD2,-CD3, or CD5. Only four of the 23 MB AML cases (17%), compared to 36% in other AML subsets, were positive for CD4 (P ¼ 0.05). B lineage antigens were also usually absent. Although most of MB AML had myeloid cytological features, no APL was observed in this subset. MC subset: The cytological features of this group are mainly those of M1 and M2 AML (47%). Nine M6 AML also belonged to this immunophenotypic subgroup. CD7 was detected in 33% of MC AML without any difference between the clinical or hematological parameters of CD7-negative and -positive groups. Conversely, different biological AML properties were observed between the two groups since APL were only observed in the CD7- subgroup, and the expression of CD7 was related to a significantly more frequent expression of B lineage antigens

(33% in CD7+ MC AML vs 12% in CD7- MC AML: Po0.004). No individual B lineage antigen was found to be specifically expressed in CD7+ MC AML. MD subset: MD AML represent 48% of the whole population. Their cytological features are heterogeneous, but M4 and M5 AML are significantly more frequent in the MD group than in the other subsets (41 vs 17%; Po105). Ten of the 24 (42%) acute erythroblastic leukemias (AEL) observed also belonged to this immunological subgroup. Significantly higher (Po0.03) median WBC counts are observed in MD AML patients compared to other AML subsets. High WBC count appears to be related to a more frequent organomegaly at presentation (P ¼ 0.01) and to CD35 or CD36 blast cell expression (Po0.007). MD AML display a significantly more frequent expression of CD14 and CD65 than other subgroups (Table 2). Both antigens are preferentially detected in CD35+ or CD36+ AML cases since CD14 and CD65 were positive in 51 and 71% of cases, respectively, in the CD35+ or CD36+ subset, compared to 21% (Po105) and 55% (Po0.002), respectively, in the CD35-/ CD36- MD AML population. CD14 positivity was also strongly related to the expression of CD11b, with 71% CD14+ MD AML cases expressing CD11b, compared to 41% in the CD14- MD AML subgroup (Po105). The expression of both CD14 and CD11b was related to high WBC counts, since 56% CD14+ and 62% CD11b+ MD AML patients had WBC 430 G/L at diagnosis compared to 36% CD14- (P ¼ 0.00007) and 44% CD11b- MD AML cases (Po0.002). In all, 88% of the CD35- or CD36positive MD AML coexpressing CD14 and CD11b had monoblastic cytologic features. AEL were also related in most cases (70%) to the expression of CD35 or CD36. However, although all the AEL were found to express CD36, within MC or ME subsets, four of 10 (40%) AEL did not express CD36 in the MD subset (Po0.03). CD71 and glycophorin A expression was detected in 83 and 70% of cases, respectively, the proportions being similar in MC and ME subsets. CD35 - and CD36- negative MD AML expressed CD34 more frequently than MD AML expressing either CD35 or CD36 or both (51 vs 41%; P ¼ 0.03). ME subset: ME AML displayed heterogeneous cytological features (Table 2). The more frequently expressed antigen was Leukemia


520 Table 2

Characteristics of AML patients of the whole population of 909 patients according to the immunological classification

subsets

n Age Median (range) Sex M/F FAB M0 M1 M2 M3 M4 M5 M6 M7 Unclassifiable WBC count (109/l) Median (range) Hb (g/dl) Median (range)

MA

MB

MC

MD

ME

179

55

180

449

46

P

59 (6–79) 0.97

55 (4–83) 0.8

59 (3–88) 1.3

54 (3–87) 1.2

53 (2–78) 1.2

NS NS

7 31 27 15 9 3 0 0 6 14 (0.8–285)

15 49 26 0 2 2 0 0 6 27 (0.7–270)

9 22 25 8 11 13 5 2 5 13.4 (0.6–856)

5 16 22 6 20 21 2 1 7 20.5 (0.5–620)

4 15 26 9 11 15 9 4 6 7.4 (0.5–116)

o10–4a o10–4a

9 (3.7–15)

9 (4–12.5)

9.3 (3.8–14.2)

9.2 (2.1–15.4)

9 (4.5–15.7)

o0.004b NS

Platelet count (109/l) Median (Range)

58 (3–560)

59 (10–324)

65 (6–400)

58 (5–556)

57 (9–422)

NS

Extramedullary disease (% of cases) Adenopathy Splenomegaly Hepatomegaly CNS involvement

26 13 12 11 2

22 11 11 6 3

34 18 23 13 0

33 19 14 12 2

41 15 23 18 3

o0.05c NS 0.04b NS NS

Immunophenotype (% of AML cases) T lineage Ag CD2 CD3 CD5 CD7

8 2 2 0

13 4 10 100

15 6 7 33

19 3 5 20

18 0 5 11

0.008b NS NS –

B lineage Ag CD10 CD19 CD20 CD22 CD24

5 7 3 4 4

0 11 0 0 6

4 14 4 6 5

6 9 5 4 12

2 16 2 5 11

NS NS NS NS 0.016b

Myeloid lineage Ag CD11b CD13 CD14 CD15 CD16 CD33 CD35 CD36 CD65 CD117

27 74 6 0 3 79 0 0 20 79

35 77 2 0 5 82 0 0 20 83

42 78 20 0 7 79 39 84 35 83

53 85 39 100 19 79 37 52 65 80

22 0 17 52 26 0 15 54 49 0

NS F o10–5a F o10–5b F F F o10–5a F

Nonlineage Ag CD34 HLA-DR TdT

39 45 12

52 72 10

53 77 19

45 67 21

41 65 21

NS o10–5d NS

a

MD vs other AML. Kruskall–Walis test; c MA and MB vs other AML. d MA vs other AML. b

myeloperoxidase (61%), and most of ME AML had an EGIL myeloid score higher than 2 (72%). Lymphoid lineage antigens were infrequently expressed. Two M7 AML expressed exclusively platelet-related antigens including CD41 and CD36. Leukemia

M6 AML expressed only glycophorin A and CD36. Five AML with monocytic differentiation expressed only CD15,CD35, or CD36. All APL were positive for CD15 but did not express CD65. Three remaining ME AML with poor antigen


o0.00003

NS o0.008

0.0002

NS 0.04

o0.01 0.0001

o0.0003 0.00001

o10–5

NS NS

NS NS

NS

o0.04 o0.03

NS

NS 0.24

47/17 5/7 29/9.5 40/14.1 33/21 60/11.5 23/10.2 38/13 30/12 38/12.2 30/9.5 1/15 o10–5 0.00002

o10

–5

0.48

o0.004

0.0001

0.0001

‘ 25/10.2 40/11.4 42/13.7 58/18 39/12.3 32/10.7 42/13.4 37/12.5 42/13 37/11 44/14 o10–5

76 53 59 73 89 65 58 70 65 68 58 74 352 147 174 327 56 442 95 406 114 384 200 297 CD34

CD14

CD7

Cytologic subtype

WBC count

*Indicates univariate analysis. **Indicates multivariate analysis. APL: Acute promyelocytic leukemia.

464 280 296 448 63 681 164 580 176 568 344 400 o60 >60 >30 o30 APL Non-APL Positive Negative Positive Negative Positive Negative Age

P** P** P* 3 years DFS (%)/ median (month) n

n

%

P

Disease-free survival CR

Response to treatment according to six prognostic factors

The prognostic factors shown to influence outcome in the overall patients population had obviously not the same weight in each immunological AML subset, since, for instance, APL were not observed in MB AML, and MA AML never expressed CD7. We therefore examined the impact of these prognostic factors in each immunological AML subset (Table 5). In univariate analysis, the response to treatment and probability of OS were related to the age of patients, and the cytological AML subtype in MA, MC, and MD AML (Table 5). WBC counts 430 G/L influenced also the achievement of CR in MC and MD AML and the probability of OS in MA, MC, and MD AML. In addition, CD34 expression on blast cells was related to a lower response rate, and shorter OS in MA and MD subsets (Figures 4a and 5a). CD34 expression influenced also the response duration in MA AML but not in MD AML (Figures 4b and 5b). CD7 positivity was associated to poor response (Po0.002) and survival (Po0.02) in MD AML. CD14 expression was related to a lower CR rate in ME AML and a shorter survival duration in MD and ME AML (Figures 6 and 7). MB AML patients had a poor outcome with a 9% probability of 3 years OS. No factor was found to influence patient outcome in this subset.

Table 3

Immunophenotypic classification-dependent prognostic factors

Overall survival

The relevance of the immunological classification on outcome was assessed on the overall population of patients who received a curative treatment. In all, 165 patients (18%) who received only supportive care were excluded from this analysis. The proportion of cases excluded from the survival analysis was similar in the different immunological AML subsets. Patients who received as consolidation treatment either chemotherapy or allogenic transplantation were also well balanced in the various immunological AML subtypes. Among the treated patients, 496 (66.7%) achieved CR and 218 (44%) had already relapsed. A total of 435 patients had died (58%) at a median time from diagnosis of 189 days. The median follow-up for alive patients was 28 months. The initial response to treatment and the probability of OS were not significantly different between the training and the test series. In univariate analysis, age460, WBC430 G/L, non-APL, blast cell expression of CD7 or CD34 were shown to display adverse prognostic significance on both the response to treatment and the OS (Table 3). CD14 detection on blast cells did not influence achievement of CR, but was related to lower OS. The immunological classification had no direct impact on CR and OS probability. Conversely, the response duration estimated by the DFS was significantly shorter in MB AML compared to remaining AML subsets (Po0.04) (Table 4). DFS was also influenced by the age of the patient and the cytological AML subgroup (Table 3). In Cox multivariate analysis, the parameters influencing survival duration remained age460 (Po106), CD34+ (Po0.00003), WBC430 G/L (Po0.0003), non-APL (Po0.01), and age460 (Po105), MB AML (Po0.04), and non-APL (Po0.04) were independent pejorative factors for DFS duration.

3 years OS (%)/ median (month)

P*

Clinical course of treated patients

o10–6

521 myeloid differentiation had myeloperoxidase cytochemical positivity.

Leukemia


522 Table 4

Response to treatment according to the immunological classification Overall survival CR

Relapse

Disease-free survival 3 years DFS %/median (month)

n

n

%

N

%

MA MB MC MD ME

153 41 141 370 39

100 28 89 252 27

65 68 63 68 69

47 17 34 105 15

47 61 38 42 55

34/12 7/9.9* 44/15.5 45/13.2 40/14.2

34/10.3 9/11.8 34/10.7 39/13.8 36/15.7

Total

744

496

67

218

44

41/13.1

35/13.6

*MB vs others: Po0.04 (logrank test).

a

1.0 CD34+ 0.9

CD34-

Probability of overall survival

0.8 0.7 0.6 0.5 0.4 0.3 p < 0.002

0.2 0.1 0.0 0

500

1000

1500

2000

2500

Survival time (days)

b

1.0 CD34+ CD34-

Probability of Disease Free Survival

0.8

0.6

0.4

0.2

p < 0.003

0.0 0

500

1000

1500

2000

Disease Free Survival time (days)

Figure 4

Leukemia

Overall survival 3 years OS %/median (month)

OS (a) and DFS (b) of patients with MA AML according to the expression of CD34.

2500

70 100 NS 9.7 17 NS

F

29 6.2

75 o0.02 18.7 o0.01

F

53 9.7

78 19.7

NS NS

F

Multivariate analysis was performed in MA, MC, and MD AML to determine the relative value of the several prognostic parameters associated with shorter OS in each of these subgroups. In MA AML, the Cox model indicated that old age (Po3 105), CD34+ (Po3 104), and WBC430 G/L (P ¼ 0.002) were independent factors pejoratively influencing the probability of OS. In MC AML, old age (Po2 105) and non-APL (Po0.03) were independent pejorative predictive factors for survival. In MD AML, multivariate analysis (po106) showed that only age460 (Po106), CD34+ (Po0.002), CD14+ (Po0.03), and WBC430 G/L (Po0.05) remained independent prognosis factors for OS. Altogether, these results show that the value on patient outcome of both immunophenotypic and nonimmunophenotypic prognosis criteria depends on the immunological differentiation of blast cells and therefore on the proposed immunological AML classification. Thus, we combined these classificationdependent prognostic factors to define, in each classification subgroup, a specific pattern of clinically relevant prognosis factors influencing the probability of OS (Figure 8).

F

75 17.7

45 9.7

0.08 NS

F

100 42

64 15

NS NS

F

Discussion

**

*

68 19 (%) (median)

Indicates univariate analysis: Indicates multivariate analysis.

NS NS

59 o0.0004 9.7 o0.002 o0.05 77 15.9 o106 78 18.2 (%) (median)

54 7.7

74 17.5 (%) (median)

67 12.6

70 NS 59 78 o104 14.7 o0.006 o0.03 10.5 16.6 o0.02 o0.0002 o0.002 68 o0.02 NS 11 54 73 9.4 15 NS o0.003 o0.02 67 13

NS NS

o105 o105

96 27

NS NS NS 55 NS 10 67 12 58 o 0.03 10 0.05 0.01 61 9.8 91 29 0.05 NS

NS 53 9 0.002 71 0.00003 2 105 11.4

68 F F NS NS 65 12.8 79 12.3 F NS NS 53 11.7 77 11.8 (%) (median)

48 6

F 65 NS o0.03 0.08 64 9 79 19 0.002 NS 0.06 53 7.7 o0,002 69 0.00002 3 105 11.7 51 6 76 13.3 (%) (median)

MA CR OS MB CR OS MC CR OS MD CR OS ME CR OS

F + P ** P* APL Non-APL P ** P* >30 o30 P ** P* o60 years >60 years

67 15.4 NS

F F 68

66 11

68

NS NS

NS NS 67 65 10.5 9.5

P* P*

P **

+

F

CD14 CD7 Cytologic subtype WBC count Age

Response to treatment and outcome of patients in each immunological AML subset according to identified prognostics factors Table 5

62 8.9

F NS NS 62 74 10.5 11.8

72 o0.02 13.6 o0.02 o3 104 53 6.7

P*

523

P **

+

F

CD34

P**


The present report demonstrates that the analysis of seven differentiation antigens, namely CD13,-CD33,-CD117,-CD7,CD35,-CD36,-CD15, on AML blast cells allows to stratify AML with a high reproducibility, in five immunological subtypes, each related to specific myeloid differentiation stages and prognostic factors. The antigens chosen for this classification are sufficient since all AML cases expressed at least one of them. The expression of the pan-myeloid antigens CD13,-CD33, and CD117 had a very close distribution in the population studied. The combination of these three markers is highly sensitive for AML diagnosis, 95% of cases being positive for at least one of them, and appears to be slightly higher than the unique detection of the cytoplasmic myeloperoxidase antigen.(10) Moreover, adding the search for MPO antigen positivity as a lineage assignment marker, more than 98% of the cases reported here could be identified as AML. Conversely, the negativity of CD13,-CD33, and CD117 defined a small group of acute leukemia that remain of myeloid origin in most cases with expression of MPO (61%) or CD65 (35%). This AML group does not usually display immaturity features, since it has a low frequency of CD7 and CD34 positivity and few cytologically undifferentiated cases were observed among ME. These AML cases rather appear to be derived from committed progenitors. Interestingly, AML cases that did not express any of the major myeloid antigens, including MPO and CD65, were all derived from the erythroblastic or megakaryoblastic lineage. According to the proposed immunological classification, AML blast cells, regardless of the FAB classification,(11) can be usefully and simply characterized by the expression of surface antigens and subsequently by their immunological differentiation. The combined analysis of CD35, CD36, and CD15 was able to provide insights about the myeloid sublineage of most differentiated AML cases. CD35 and CD36 are usually detected on many normal myeloid lineage cells including erythroid precursors and monocytes.(12,13) CD36 analysis was sensitive to detect erythroblastic lineage-derived AML (82%), and the only four M6 cases that did not express CD36 belonged to the MD subset. The coexpression of CD35-or-CD36 with CD15 identifies most of the AML with a monocytic component, which also Leukemia


524

a

1.0 CD34+

0.9

CD34-

Probability of Overall Survival

0.8 0.7 0.6 0.5 0.4 0.3 p = 0.015 0.2 0.1 0.0 0

500

1000

1500

2000

2500


b

1.0 CD34+

0.9

CD34Probability of Disease Free survival

0.8 0.7 0.6 0.5 p = 0.9 0.4 0.3 0.2 0.1 0.0 0

500

1000

1500

2000

2500

Disease Free Survival time (days)

Figure 5

OS (a) and DFS (b) of patients with MD AML according to the expression of CD34.

usually express CD14 and CD11b. Therefore, CD15, which has been reported to be a late differentiation antigen,(14) is also strongly expressed in monoblastic AML populations. Conversely, CD15-positive AML lacking CD35 and CD36 appeared to be mainly of granulocytic lineage. In all, 53% of AML cytologically defined as myeloid (M1–M3) were characterized by early myeloid differentiation (CD13,CD33, or CD117 positivity), minimal terminal granulocytic differentiation without expression of CD15,-CD35, or CD36. They were classified as MA or MB according to the expression of CD7. The distance linkage between CD7 and the cluster CD13/ CD33/CD117 was larger than that observed between CD34 and the pan-myeloid antigens. Moreover, the frequency of CD7 expression decreased along granulomonocytic AML maturation from MB to ME (2), while the frequency of CD34 positivity was quite similar in all these AML subsets. This indicates that, as in Leukemia

normal myeloid differentiation,(15,16) CD7 is likely to be expressed in a small compartment of immature cells, while CD34 is expressed on a larger population of both immature and committed myeloid cells. Altogether these data led us to select CD7 rather than CD34 as an immaturity marker to establish the classification. They also suggest that MB AML cases, coexpressing CD7 and only immature pan-myeloid antigens, without evidence of granulomonocytic terminal differentiation, could be derived from an immature myeloid precursor found in a subset of bone marrow and fetal liver myeloid progenitors.(15–17) Consistent with this hypothesis, the percentage of CD7+ AML blast cells was significantly higher in the MB AML subset compared to other AML subsets (Po0.05; data not shown). It has been suggested that the immunophenotype of the blastic cell population may depend on the differentiation ability of the leukemia-initiating cell.(18) Subsequently, the biological


525 1.0 CD14+

0.9

CD14-

Probability of Overall Survival

0.8 0.7 0.6 0.5 0.4 0.3 p = 0.035

0.2 0.1 0.0 0

Figure 6

500

1000 1500 Survival time (days)

2000

2500

OS of patients with MD AML according to the expression of CD14.

1.0

CD14+ CD14-

0.9

Probability of overall survival

0.8 0.7 0.6 0.5 0.4 0.3 p < 0.03

0.2 0.1 0.0 0

500

1000

1500

2000

2500


Figure 7

OS of patients with ME AML according to the expression of CD14.

significance of the expression of a given differentiation antigen will vary depending on the AML immunological subset considered, yielding variable prognostic values. This has important bearings as the presence of a given marker on blast cells should no longer be considered individually, but should be replaced in the context of myeloid differentiation according to the immunological classification of AML proposed here. This report highlights specifically the fact that the prognostic value of an individual antigen’s expression depends on the immunological differentiation stage of the leukemic cells. The present study demonstrated that antigens whose expression had already been shown as a potential prognostic factor (CD34,-CD7,-

CD14)(19–28) were indeed related to outcome in the overall population. More interestingly, their prognostic value, which could be marginal in the overall population, became strongly significant in specific immunological AML subsets while it retained no significance in other subsets. The prognostic influence of CD7 positivity thus appears to be restricted to the MB subset. This had already been pointed out indirectly by reports analyzing infrequent AML cases expressing either CD13 or CD33 and CD7 without other T or myeloid lineage antigen.(19, 29) Conversely, the influence of CD7 expression on CR achievement in the overall AML population was not related, in multivariate analysis, to a shorter DFS or OS of CD7+ AML Leukemia


526 CD13 or CD33 or CD117

CD7

CD35 or CD36

CD15

CD7

Old Age

Non Immunologic criteria High WBC APL/non APL

MA

+

-

-

-

-

-

+

+

+

-

MB

+

+

-

-

+

-

-

-

-

-

MC

+ +

+/-

+

-

-

-

-

+

-

+

MD

+

+/-

+/-

+

-

+

+

+

+

-

ME

-

+/-

+/-

+/-

-

+

-

-

-

-

Figure 8

Immunophenotypic classification-dependent prognostic parameters influencing pejoratively the overall survival duration.

patients probably because of the overlapping prognostic influence of APL. The CD7 negativity of APL(6) had not previously been accounted for in studies reporting a poor survival of CD7+AML.(20,25) Yet antigens never expressed by some specific progenitor-derived AML blast cells could obviously not have any prognostic value in these AML subsets. CD34 expression was related to poor outcome in 70% of AML of the MA and MD AML subsets. The frequency of CD34 expression was similar in these two subsets as compared to other immunological AML subsets, and was not related to specific FAB subgroups, which indicates that MA and MD indeed define specific entities. Conversely, CD14, more frequently expressed in AML with a monocytic component,(21–23,30) appeared related to a poor outcome only in MD and ME subsets. The prognostic value of CD14 expression was independent of the FAB classification in the MD subset and had already been reported in such FAB subsets as M3(22) or M0.(2) The imbalanced prognostic weight of CD7,-CD14, or CD34 expression in different immunological settings could explain the discrepancies previously observed when analyzing individually the prognostic value of these antigens expression.(19–21, 24–28, 30–39) The relation between the proposed immunological classification of AML and cytogenetic subsets remain to be determined. We previously demonstrated that the expression of a single given antigen on blast cells was poorly related to karyotype, and whether some immunophenotypic patterns were strongly predictive of cytogenetic entities, numerous immunophenotypic variants were observed among AML subtypes with recurrent cytogenetic anomalies.(6) This AML classification could help to identify some of these clinically relevant variants. It is particularly interesting to note that APL cases displaying an immature phenotype (MA) have a poorer clinical outcome than that of patients with more mature APL (MC to ME): MA APL patients achieve less frequently CR (82 vs than 97%) (P ¼ 0.04) and have a worse 3-year OS probability (54 vs 68% although statistical significance was not reached at the time of testing, P ¼ 0.09), compared to mature APL patients (data not shown). Further analyses will be needed to confirm these data on a larger number of APL cases and to determine the value of this immunophenotypic classification to stratify AML populations with either other recurrent cytogenetic anomalies or normal karyotypes. From a practical point of view, the seven markers allowing to apply this classification should be completed by the investigation of the MPO antigen, in order to confirm in all cases that blast cells belong to the myeloid lineage, and by that of CD14and-CD34 for prognostic purposes. Megakaryocytic markers such as CD41 or CD61 should also be incorporated into the immunophenotyping panel, as well as CD65-and-CD11b for a more complete identification of the differentiation stage. Leukemia

Immunologic criteria CD14 CD34

In summary, the immunological classification of AML that we propose, validated on a large cohort of patients, allows to easily identify five clinically relevant AML subsets with specific differentiation characteristics. In each classification subset, high-risk AML patients can be identified, which should be eligible for alternative therapeutic strategies.

References 1 Bene MC, Castoldi G, Knapp W, Ludwig WD, Matutes E, Orfao A et al. Proposals for the immunological classification of acute leukemias. European Group for the Immunological Characterization of Leukemias (EGIL). Leukemia 1995; 9: 1783–1786. 2 Bene MC, Bernier M, Casasnovas RO, Castoldi G, Doekharan D, van der Holt B et al. Acute myeloid leukaemia M0: haematological, immunophenotypic and cytogenetic characteristics and their prognostic significance: an analysis in 241 patients. Br J Haematol 2001; 113: 737–745. 3 Matutes E, Morilla R, Farahat N, Carbonell F, Swansbury J, Dyer M et al. Definition of acute biphenotypic leukemia. Haematologica 1997; 82: 64–66. 4 Garand R, Vannier JP, Bene MC, Faure G, Favre M, Bernard A. Comparison of outcome, clinical, laboratory, and immunological features in 164 children and adults with T-ALL. The Groupe d’Etude Immunologique des Leucemies. Leukemia 1990; 4: 739– 744. 5 Garand R, Voisin S, Papin S, Praloran V, Lenormand B, Favre M et al. Characteristics of pro-T ALL subgroups: comparison with late T-ALL. The Groupe d’Etude Immunologique des Leuce´mies. Leukemia 1993; 7: 161–167. 6 Casasnovas RO, Campos L, Mugneret F, Charrin C, Be´ne´ MC, Garand R et al. Immunophenotypic patterns and cytogenetic anomalies in acute non-lymphoblastic leukemia subtypes: a prospective study of 432 patients. Leukemia 1998; 12: 34–43. 7 Hartigan JA. Clustering algorithms. New York: Wiley, 1975. 8 Sneath PHA, Sokal RR. Numerical Taxonomy. San Francisco: WH Freeman & Co., 1973. 9 Bene MC, Bernier M, Casasnovas RO, Castoldi G, Knapp W, Lanza F et al. The reliability and specificity of c-kit for the diagnosis of acute myeloid leukemias and undifferentiated leukemias. The European Group for the Immunological Classification of Leukemias (EGIL). Blood 1998; 92: 596–599. 10 Buccheri V, Shetty V, Yoshida N, Morilla R, Matutes E, Catovsky D. The role of an anti-myeloperoxidase antibody in the diagnosis and classification of acute leukaemia: a comparison with light and electron microscopy cytochemistry. Br J Haematol 1992; 80: 62– 68. 11 Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gralnick HR et al. Proposals for the classification of the acute leukaemias. French-American-British (FAB) co-operative group. Br J Haematol 1976; 33: 451–458. 12 Huh HY, Pearce SF, Yesner LM, Schindler JL, Silverstein RL. Regulated expression of CD36 during monocyte-to-macrophage differentiation: potential role of CD36 in foam cell formation. Blood 1996; 87: 2020–2028.


527 13 van Schravendijk MR, Handunnetti SM, Barnwell JW, Howard RJ. Normal human erythrocytes express CD36, an adhesion molecule of monocytes, platelets, and endothelial cells. Blood 1992; 80: 2105–2114. 14 Lo SK, Golenbock DT, Sass PM, Maskati A, Xu H, Silverstein RL. Engagement of the Lewis X antigen (CD15) results in monocyte activation. Blood 1997; 89: 307–314. 15 Tien HF, Chou CC, Wang CH, Chang CH, Hsing CC. Putative normal counterparts of leukaemic cells from CD7-positive acute myeloid leukaemia can be demonstrated in human haemopoietic tissues. Br J Haematol 1996; 94: 501–506. 16 Chabannon C, Wood P, Torok-Storb B. Expression of CD7 on normal human myeloid progenitors. J Immunol 1992; 149: 2110– 2113. 17 Barcena A, Muench MO, Galy AH, Cupp J, Roncarolo MG, Phillips JH et al. Phenotypic and functional analysis of T-cell precursors in the human fetal liver and thymus: CD7 expression in the early stages of T- and myeloid-cell development. Blood 1993; 82: 3401–3414. 18 Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med 1997; 3: 730–737. 19 Jensen AW, Hokland M, Jorgensen H, Justesen J, Ellegaard J, Hokland P. Solitary expression of CD7 among T-cell antigens in acute myeloid leukemia: identification of a group of patients with similar T-cell receptor beta and delta rearrangements and course of disease suggestive of poor prognosis. Blood 1991; 78: 1292–1300. 20 Venditti A, Del Poeta G, Buccisano F, Tamburini A, Cox-Froncillo MC, Aronica G et al. Prognostic relevance of the expression of Tdt and CD7 in 335 cases of acute myeloid leukemia. Leukemia 1998; 12: 1056–1063. 21 Solary E, Casasnovas RO, Campos L, Bene MC, Faure G, Maingon P et al. Surface markers in adult acute myeloblastic leukemia: correlation of CD19+, CD34+ and CD14+/DR–phenotypes with shorter survival. Groupe d’Etude Immunologique des Leucemies (GEIL). Leukemia 1992; 6: 393–399. 22 Fergedal M, Astrom M, Tidefelt U, Karlsson MG. Differences in CD14 and alpha-naphthyl acetate esterase positivity and relation to prognosis in AML. Leuk Res 1998; 22: 25–30. 23 Campos L, Guyotat D, Archimbaud E, Devaux Y, Treille D, Larese A et al. Surface marker expression in adult acute myeloid leukaemia: correlations with initial characteristics, morphology and response to therapy. Br J Haematol 1989; 72: 161–166. 24 Bradstock K, Matthews J, Benson E, Page F, Bishop J. Prognostic value of immunophenotyping in acute myeloid leukemia. Australian Leukaemia Study Group. Blood 1994; 84: 1220–1225. 25 Del Poeta G, Stasi R, Venditti A, Cox C, Aronica G, Masi M et al. CD7 expression in acute myeloid leukemia. Leuk Lymphoma 1995; 17: 111–119. 26 Del Poeta G, Stasi R, Venditti A, Suppo G, Aronica G, Bruno A et al. Prognostic value of cell marker analysis in de novo acute myeloid leukemia. Leukemia 1994; 8: 388–394.

27 Lee EJ, Yang J, Leavitt RD, Testa JR, Civin CI, Forrest A et al. The significance of CD34 and TdT determinations in patients with untreated de novo acute myeloid leukemia. Leukemia 1992; 6: 1203–1209. 28 Geller RB, Zahurak M, Hurwitz CA, Burke PJ, Karp JE, Piantadosi S et al. Prognostic importance of immunophenotyping in adults with acute myelocytic leukaemia: the significance of the stem-cell glycoprotein CD34 (My10). Br J Haematol 1990; 76: 340–347. 29 Bassan R, Biondi A, Benvestito S, Tini ML, Abbate M, Viero P, Barbui T et al. Acute undifferentiated leukemia with CD7+ and CD13+ immunophenotype. Lack of molecular lineage commitment and association with poor prognostic features. Cancer 1992; 69: 396–404. 30 Merle-Beral H, Nguyen Cong Duc L, Leblond V, Boucheix C, Michel A et al. Diagnostic and prognostic significance of myelomonocytic cell surface antigens in acute myeloid leukaemia. Br J Haematol 1989; 73: 323–330. 31 Schwarzinger I, Valent P, Koller U, Marosi C, Schneider B, Haas O Prognostic significance of surface marker expression on blasts of patients with de novo acute myeloblastic leukemia. J Clin Oncol 1990; 8: 423–430. 32 Miwa H, Nakase K, Kita K. Biological characteristics of CD7(+) acute leukemia. Leuk Lymphoma 1996; 21: 239–244. 33 Saxena A, Sheridan DP, Card RT, McPeek AM, Mewdell CC, Skinnider LF. Biologic and clinical significance of CD7 expression in acute myeloid leukemia. Am J Hematol 1998; 58: 278–284. 34 Kornblau SM, Thall P, Huh YO, Estey E, Andreeff M. Analysis of CD7 expression in acute myelogenous leukemia: martingale residual plots combined with ’optimal’ cutpoint analysis reveals absence of prognostic significance. Leukemia 1995; 9: 1735– 1741. 35 Reuss Borst MA, Bu¨hring HJ, Schmidt H, Mu¨ller CA. AML: immunophenotypic heterogeneity and prognostic significance of c-kit expression. Leukemia 1994; 8: 258–263. 36 Kanda Y, Hamaki T, Yamamoto R, Chizuka A, Suguro M, Matsuyama T et al. The clinical significance of CD34 expression in response to therapy of patients with acute myeloid leukemia: an overview of 2483 patients from 22 studies. Cancer 2000; 88: 2529–2533. 37 Kyoda K, Nakamura S, Hattori N, Takeshima M, Nakamura K, Kaya H et al. Lack of prognostic significance of CD34 expression in adult AML when FAB M0 and M3 are excluded. Am J Hematol 1998; 57: 265–266. 38 Lanza F, Rigolin GM, Moretti S, Latorraca A, Castoldi G. Prognostic value of immunophenotypic characteristics of blast cells in acute myeloid leukemia. Leuk Lymphoma 1994; 13(Suppl 1): 81–85. 39 Ciolli S, Leoni F, Caporale R, Pascarella A, Salti F, Rossi-Ferrini P. CD34 expression fails to predict the outcome in adult acute myeloid leukemia. Haematologica 1993; 78: 151–155.

Leukemia