ORIGINAL ARTICLE. A van Rhenen 1, B Moshaver 1, A Kelder 1, N Feller 1, AWM Nieuwint 2, S Zweegman 1, GJ Ossenkoppele 1 and GJ Schuurhuis 1

(2007) 21, 1700 1707 & 2007 Nature Publishing Group All rights reserved 0887-6924/07 $30.00 www.nature.com/leu ORIGINAL ARTICLE Aberrant marker expression patterns on the CD34 þ CD38 stem cell compartment in acute myeloid leukemia allows to distinguish the malignant from the normal stem cell compartment both at diagnosis and in remission A van Rhenen 1, B Moshaver 1, A Kelder 1, N Feller 1, AWM Nieuwint 2, S Zweegman 1, GJ Ossenkoppele 1 and GJ Schuurhuis 1 1 Department of Hematology, VU University Medical Center, Amsterdam, The Netherlands and 2 Department of Cytogenetics, VU University Medical Center, Amsterdam, The Netherlands Acute myeloid leukemia (AML) is generally regarded as a stem cell disease. In CD34-positive AML, the leukemic stem cell has been recognized as CD38 negative. This CD34 þ CD38 population survives chemotherapy and is most probable the cause of minimal residual disease (MRD). The outgrowth of MRD causes relapse and MRD can therefore serve as a prognostic marker. The key role of leukemogenic CD34 þ CD38 cells led us to investigate whether they can be detected under MRD conditions. Various markers were identified to be aberrantly expressed on the CD34 þ CD38 population in AML and high-risk MDS samples at diagnosis, including C-type lectin-like molecule-1 and several lineage markers/marker-combinations. Fluorescent in situ hybridization analysis revealed that marker-positive cells were indeed of malignant origin. The markers were neither expressed on normal CD34 þ CD38 cells in steady-state bone marrow (BM) nor in BM after chemotherapy. We found that these markers were indeed expressed in part of the patients on malignant CD34 þ CD38 cells in complete remission, indicating the presence of malignant CD34 þ CD38 cells. Thus, by identifying residual malignant CD34 þ CD38 cells after chemotherapy, MRD detection at the stem cell level turned out to be possible. This might facilitate characterization of these chemotherapy-resistant leukemogenic cells, thereby being of help to identify new targets for therapy. (2007) 21, 1700 1707; doi:10.1038/sj.leu.2404754; published online 24 May 2007 Keywords: AML; CD34 þ CD38 ; MRD; stem cell markers Introduction In acute myeloid leukemia (AML), complete remission can be achieved using intensive chemotherapy in the majority of patients, but relapse is common. Relapse likely originates from minimal residual disease (MRD) cells, which can be detected after chemotherapy. Consequently, quantifying MRD in AML, using immunophenotypical or molecular procedures, is of prognostic importance: MRD frequency after chemotherapy, determined by flow cytometry, strongly correlates with the incidence of relapse and survival. 1 4 AML is generally regarded as a stem cell disease and in CD34- positive AML the stem cell has been recognized as CD38 negative. 5 These CD34 þ CD38 cells have chemotherapyresistant properties 6 and are therefore likely responsible for the outgrowth of MRD, which in turn is thought to cause relapse. Indeed, we have shown that a high leukemic CD34 þ CD38 Correspondence: Dr GJ Schuurhuis, Department of Hematology, VU University Medical Center, CCA 4.24, De Boelelaan 1117, 1081 HV Amsterdam, The Netherlands. E-mail: gj.schuurhuis@vumc.nl Received 14 November 2006; accepted 23 April 2007; published online 24 May 2007 stem cell frequency at diagnosis significantly correlates with both high MRD frequency after chemotherapy and short survival. 7 MRD frequency can be determined by flow cytometry using aberrant antigen expression defined on the leukemic blasts at diagnosis. These abnormal immunophenotypes (leukemiaassociated phenotypes, LAPs) consist of aberrant cross-lineage marker expression, asynchronous expression and increased or decreased marker expression. 1 4 We hypothesized that these abnormal immunophenotypes, that are defined on the malignant progenitor population, might also be present on the leukemic CD34 þ CD38 stem cell compartment which would enable the detection of the malignant CD34 þ CD38 stem cell compartment after chemotherapy. If so, assessing the frequency of these malignant CD34 þ CD38 cells may provide additional prognostic information. Moreover, it would make the leukemic stem cell compartment accessible for functional or molecular studies under remission conditions, which may be of great importance for the design of therapies directed against such chemotherapy-resistant leukemic stem cells. 8 In addition to these lineage-related aberrancies, other markers present on AML blasts, are available to discriminate malignant CD34 þ CD38 cells in remission bone marrow (BM). We recently described that C-type lectin-like molecule-1 (CLL-1) is present on AML CD34 þ CD38 cells in the majority of patients and absent on CD34 þ CD38 cells in normal BM (NBM), which makes it a suitable stem cell marker (van Rhenen et al., Blood (2005) 106, 6a abstract). Another potential candidate is the interleukin-3-receptor a-chain (CD123), which has also been described as a marker for AML CD34 þ CD38 cells. 9 In this study, using the above-mentioned markers, we investigated the possibilities for the detection of leukemic CD34 þ CD38 cells in complete remission. NBM and regenerating BM (RBM) samples served as controls. Follow-up analyses were performed in samples of AML patients with a detectable CD34 þ CD38 population with MRD frequency (further referred to as whole blast MRD) detected in parallel. We show that in AML and in high-risk MDS, CD34 þ CD38 cells are at diagnosis in the majority of the cases characterized by one or more aberrancies. These aberrancies allow to discriminate normal and AML CD34 þ CD38 cells in MRD, thereby offering possibilities for applications and perspectives mentioned earlier. Materials and methods Patient samples BM samples of 55 patients presenting with AML or high-risk MDS at the VU University Medical Center were obtained after

informed consent at diagnosis and after chemotherapeutic treatment. In five cases, BM was not available at diagnosis and peripheral blood was used. BM infiltration of the high-risk MDS patients was median 15% (range 5 28%). After chemotherapeutic treatment only BM samples were used. Diagnosis of patients was based on morphology using the French- American-British (FAB) classification, immunophenotyping, molecular biology and cytogenetics. 10 Patient characteristics are shown in Table 1. NBM was obtained after informed consent from patients undergoing cardiac surgery. RBM was obtained from five patients with acute lymphoblastic leukemia, from one patient with non-hodgkin s lymphoma, from three patients with CD34-positive AML but absence of expression of either CLL-1 or lineage markers, and from four patients with CD34-negative AML (o1% CD34-positive cells at diagnosis). AML patients were treated according to the HOVON 42 and 43 AML studies (www.hovon.nl). The majority of the samples was analyzed freshly. Red blood cells were lysed using a 10 min lysing procedure on ice with 10 ml lysis buffer (155 mm NH 4 Cl, 10 mm KHCO 3, 0.1 mm Na 2 ethylenediaminetetraacetic acid (EDTA), ph 7.4) and washed with phosphate-buffered saline (PBS) containing 0.1% bovine serum albumin (BSA). Frozen samples were prepared using a Ficoll gradient (1.077 g/ml, Amersham Biosciences, Freiburg, Germany) and subsequent red blood cell lysis. Cells were then frozen in RPMI (Gibco, Paisley, UK) with 20% heat-inactivated fetal bovine serum (FBS, Greiner, Alphen a/d Rijn, The Netherlands) and 10% dimethylsulfoxide (Riedel-de Haen, Seelze, Germany) in isopropanol-filled containers and subsequently stored in liquid nitrogen. When needed for analysis, cells were thawed and suspended in prewarmed RPMI with 40% FBS at 371C. Cells were washed and enabled to recover for 45 min in the same medium at 371C. Cells were washed again and suspended in PBS with 0.1% BSA. Table 1 Patient characteristics No. of patients 60 Male/female 35/25 Age at diagnosis, years, median (range) 58 (17 80) WBC count at diagnosis, 109/l, median (range) 4.0 (1 322) FAB classification n (%) M0 1 (1.7) M1 7 (11.7) M2 8 (13.3) M3 3 (5.0) M4 4 (6.6) M5 10 (16.7) M6 7 (11.7) M7 0 (0.0) Refractory anemia with excess blasts 13 (21.7) Refractory anemia with excess blasts in transformation 2 (3.3) Not classified 5 (8.3) Cytogenetic risk group n (%) Favorable 10 (16.7) Intermediate 38 (63.3) Poor 7 (11.7) No metaphases 5 (8.3) Flt3 ITD n (%) Present 13 (21.7) Absent 45 (75.0) Not analyzed 2 (3.4) Abbreviation: FAB, French-American-British; ITD, internal tandem duplication; WBC, white blood cells. FACS analysis Procedures have been described before. 7 Monoclonal antibody combinations consisted of fluorescein isothiocyanate (FITC), phycoerythrin (PE), peridinyl chlorophyllin (PerCP), allophycocyanin (APC) and phycoerythrin-cy7 (PE-Cy7) labelled monoclonal antibodies. Anti-CD45 PerCP, anti-cd38 APC, anti-cd34 FITC, anti-cd34 PE-Cy7, anti-cd2 PE, anti-cd7 PE, anti-cd11b PE, anti-cd13 PE, anti-cd19 PE, anti-cd22 PE, anti-cd33 FITC, anti-cd33 PE, anti-cd56 PE, anti-cd123 PE and human leukocyte antigen (HLA)-DR FITC were all from BD Biosciences (San Jose, CA, USA). Anti-CD5 PE, anti-cd11c FITC and anti-cd13 FITC were from DAKO (Heverlee, Belgium). Anti-CLL-1 and isotype control were as described previously. 11 Data acquisition was performed using a FACScalibur equipped with an argon and red diode laser and a FACScanto (both BD Biosciences) equipped with solid-state lasers (red, blue and violet); analysis was performed using Cellquest (BD Biosciences). The gating strategy used takes into account that the CD34 þ CD38 population is a minor population and has been described before. In short, CD34 þ blasts were identified in a population characterized by dim expression of CD45 and low sideward scatter (SSC), 12 and within this CD34 þ population, the CD34 þ CD38 compartment was defined. The required minimal number of CD34 þ CD38 events was set at 20 and these events should tightly cluster in a FSC-SSC plot. PBS was used as a negative control for both CD38 and lineage marker expression, as has been validated before when comparing PBS with isotype controls for MRD detection. 1,7 LAPs were defined at diagnosis as shown previously. In this study a progenitor marker, CD34 or CD117, was combined with CD45 to define the progenitor population. These results for the whole blast compartment are further referred to as whole blast MRD. For stem cell-specific analysis with CLL-1, CD123 or lineage markers/marker combinations, anti-cd38 APC was added to a combination of anti-cd45 PerCP and anti-cd34 FITC or PE-Cy7. Combinations that were used are depicted in Table 2. Using four-color fluorescence-activated cell sorting (FACS) analysis, lineage marker expression (CD2, CD5, CD7, CD11b, CD11c, CD19 and CD22 and CD56) was studied. These single markers are almost completely absent on normal CD34 þ CD38 cells and therefore do not need addition of extra myeloid markers as far as the CD34 þ CD38 population is concerned (results of this will be discussed in Table 5). Fivecolor FACS analysis was performed for aberrancies defined by two markers and included antibody combinations for CD13, CD33 and HLA-DR expression. Table 2 Marker combinations used for flow cytometry FITC PE PE Cy7 CD45 PerCP CD38 APC N (patients) CD34 CD2 + + 3 CD34 CD5 + + 1 CD34 CD7 + + 17 CD34 CD11b + + 4 CD34 CD19 + + 2 CD34 CD56 + + 7 CD34 CLL-1 + + 43 a CD34 CD123 + + 23 a CD13 CD33 CD34 + + 7 CD33 CD13 CD34 + + 2 CD11c CD11b CD34 + + 4 HLADR CD33 CD34 + + 4 a Expression450%. 1701

1702 In most AML cases at diagnosis the complete CD34 þ CD38 population showed a shift towards positivity compared to the PBS control upon staining for aberrant marker/marker combinations, indicating a homogeneous malignant population. However, in some cases small marker-negative subpopulations were present, that likely consisted of normal CD34 þ CD38 cells as was shown in some cases by fluorescent in situ hybridization (FISH) experiments (results will be discussed in Table 3). Expression of CLL-1 and CD123 on CD34 þ CD38 cells was scored as o50% or 450% (not including the small markernegative subpopulations). Only AML samples with CD3441% were included, since we have previously shown that in samples with CD34 o1% the CD34 þ cells are in general of normal origin. 13 After the different courses of chemotherapy, follow-up analysis was performed, in which for each antigen combination at least 1 10 6 mononuclear cells were tested. FISH analysis of FACS-sorted CD34 þ CD38 cells Patient samples were FACS-sorted after Ficoll isolation, red blood cell lysis and cryopreservation. Lineage marker-positive CD34 þ CD38 populations were defined compared to a PBS control. Via-Probe (7-amino actinomycin D; BD Pharmingen, Table 3 Expression of CLL-1, CD123 and lineage markers/marker combination on CD34+CD38 cells in AML patients at diagnosis No. of patients 60 a 60 a Reliable number of CD34+CD38 events 56 56 CLL-1 b 23 23 CD123 b 43 Not included Lineage marker/marker combination 44 44 None 3 10 Total number of patients 53 46 Abbreviations: AML, acute myeloid leukemia; CLL, C-type lectin-like molecule-1. a First column including CD123. Second column excluding CD123. b Expression450%. San Diego, CA, USA) was used to exclude dead cells. Cytospins were made at 4.5G for 20 min. FISH analysis for t(8;21) and trisomy 8 was performed using LSI ETO/AML1 or the centromere probe for chromosome 8 from Vysis (IL, USA). Slides were incubated for 60 min at 371C in 100 mg RNAse A (Boehringer Mannheim, Germany)/ml 2 standard saline citrate (SSC) and washed three times in 2 SSC at 371C. Next, slides were incubated for 10 min at 371C in 0.01% pepsine (Sigma-Aldrich, St Louis, MO, USA) in 10 mm HCl, and washed two times for 5 min in PBS at room temperature (RT). Afterwards, the slides were incubated for 10 min in 3.7% formaldehyde in PBS at RT and washed two times for 5 min. Denaturation occurred at 721C, for 5 min in 70% formamide. Next, dehydration was performed in 70% EtOH at 201C two times for 5 min, in 96% EtOH for 5 min at RT and finally 5 min in 100% EtOH at RT. Slides were air-dried. Hybridization with the probes took place overnight in a wet chamber at 371C. Afterwards, slides were washed in 50% formamide/2 SSC at 421C three times for 10 min. Next slides were washed at 421C, for 10 min in 2 SSC and finally washed in 2 SSC/0.05% Tween at 421C for 5 min. Slides were dehydrated in ethanol 70, 96 and 100% for 1 min at RT and incubated for 8 min in DAPI (Sigma-Aldrich) /PBS (0.1 mg/ml) at RT. After this, slides were again dehydrated and air-dried. Vectashield (Vector Laboratories, Brunschwig, The Netherlands) was applied to the slides before analysis. Results CD34 þ CD38 populations in AML at diagnosis and relapse CD34 þ CD38 cells in AML and high-risk MDS samples at diagnosis were analyzed for the expression of CLL-1, CD123 and the markers/marker combinations derived from LAPs. Examples of FACS analyses for different patients are shown in Figures 1 and 2. Results are summarized in Table 3. A reliable number of CD34 þ CD38 events (420) could be measured in 56/60 CD34-positive cases, with a median of 293 events for CLL-1 expression, 449 events for CD123 expression and 776 events for lineage marker/marker combination expression. CLL- 1 expression was 450% in 23/60 cases, CD123 expression was Figure 1 Examples of marker defined malignant CD34 þ CD38 populations at diagnosis and in NBM. After labelling of the cells with the appropriate antibody combinations, the CD34 þ CD38 cells were identified by gating on the CD45dim/SSClow population and CD34 þ expression using a Boolean gating strategy. All data shown are within this compartment. The left and right lower quadrants represent the CD34 þ CD38 population. These examples were obtained from four different patients. (a) CD7 expression, (b) CD11b expression, (c) CD2 expression and (d) CD56 expression. (e h) Expression of the same antigens in NBM is shown. Note the presence of marker-negative cells in (a d), which represent putative normal cells. APC, allophycocyanin; PE, phycoerythrin.

cells; the lineage marker-negative population in these three cases was a distinct subpopulation with low frequency. Both sorted populations were scored for the cytogenetic abnormality (Table 4). Patient 1 had AML with t(8;21) and CD19 as lineage marker, which was present on 91% of the CD34 þ CD38 population. Patient 2 also had t(8;21) with CD56 as lineage marker, present on 77% of the CD34 þ CD38 cells. Finally, patient 3 had a leukemia with trisomy 8 (only 3/20 cells in routine cytogenetic analysis) and CD7 expression as aberrancy present on 85% of the CD34 þ CD38 population. FISH analysis indicated a malignant origin of the lineage markerpositive CD34 þ CD38 cells, whereas the corresponding lineage marker-negative CD34 þ CD38 cells did not show any or a low frequency of the cytogenetic abnormality. In conclusion, aberrant marker-positive CD34 þ CD38 cells at diagnosis are apparently of malignant origin, while the corresponding marker-negative population is likely of normal origin. 1703 Figure 2 Examples of marker combination defined malignant CD34 þ CD38 populations at diagnosis and in NBM. After labelling of the cells with the appropriate antibody combinations, the CD34 þ CD38 cells were identified by gating on the CD45dim/ SSClow population, CD34 þ expression and CD38 expression using a Boolean gating strategy. All data shown are within this compartment. Parts (a c) are from different AML patients and (d) is from NBM. (a) CD13 CD33 þ expression, (b) CD13 þ CD33 expression and (c) CD13 overexpression on the CD34 þ CD38 cells. (d) CD13 and CD33 expression on NBM CD34 þ CD38 cells. FITC, fluorescein isothiocyanate; PE, phycoerythrin. 450% in 43/60 cases. Lineage marker expression was present in all 44 cases analyzed. This is close to the number of samples that could be analyzed for whole blast MRD (47/56 samples). The only aberrant phenotype that could by definition not be analyzed on CD34 þ CD38 cells is defined on the CD34 population (CD13 þ CD34 CD117 þ ). For all cases studied, the expression on the CD34 þ CD38 population was at least as high as the expression found on the progenitor compartment (CD34 þ CD38 þ ). A first indication of the stability of CLL-1, CD123 and lineage marker expressions during treatment/relapse was obtained by comparing relapse with diagnosis samples from the same patients. In the four patients with CLL-1 expression 450% at diagnosis who suffered from a relapse, CLL-1 expression was not different at relapse. Two samples of AML patients that did show very low CLL-1 expression at diagnosis had a corresponding low expression at relapse, showing that it is unlikely that CLL-1 expression on CD34 þ CD38 cells is upregulated during treatment. Similar to CLL-1, both CD123 expression (n ¼ 6 relapsed patients) and lineage marker expression (n ¼ 2 relapsed patients) did not change at relapse. FISH analysis of marker defined CD34 þ CD38 populations in AML at diagnosis To confirm the malignant origin of marker-positive CD34 þ CD38 cells, CD34 þ CD38 populations from three patients at diagnosis with FISH-detectable cytogenetic abnormalities were FACS-sorted. All patients had aberrant lineage marker expression on the large majority of the CD34 þ CD38 CD34 þ CD38 cells in NBM Reliable detection of malignant stem cells in the course of the disease using CLL-1, CD123 or the different lineage markers/ marker combinations requires that normal stem cells do not show any or very low expression of these markers. Therefore, control BM samples from volunteers were analyzed (see Figures 1 and 2 and Table 5). CLL-1 was found to be virtually absent on normal CD34 þ CD38 cells (median expression 0%, n ¼ 10). CD123 expression was present on part of the CD34 þ CD38 stem cell population, as a result of a shift of the complete CD34 þ CD38 population towards positivity compared to the PBS control. This population showed low intensity and had a median expression of 12.0% (n ¼ 6). The expressions of the lineage markers (CD2, CD5, CD7, CD19, CD11b, CD22 and CD56) and combinations characterized by asynchronous expression (CD13 CD33 þ, CD13 þ CD33 ) were all very low on normal CD34 þ CD38 cells, with the exception of CD11c that had weak expression. Overexpression of CD13 (CD13 þþ) and CD33 (CD33 þþ) does not occur in NBM (see Figure 2d). In contrast, HLA-DR expression is very high in NBM, which makes the absence of HLA-DR aberrant. Summarizing, both CLL-1, lineage marker expression, lineage marker asynchronous antigen expression, lineage marker antigen overexpression and underexpression and CD123 are suitable for the specific detection of the malignant CD34 þ CD38 compartment in AML at diagnosis. CD34 þ CD38 populations in RBM Using stem cell markers for follow-up evaluation of the malignant CD34 þ CD38 cells in AML patients requires the opposite expression in NBM CD34 þ CD38 cells present after chemotherapeutic treatment. To check this BM samples from patients with different hematological diseases obtained after chemotherapeutic treatment were studied. Results obtained thus far with markers expressed on AML CD34 þ CD38 cells but not on normal CD34 þ CD38 cells are shown in Table 5. Similar to NBM, in RBM there was virtually no expression on CD34 þ CD38 cells of both CLL-1 and the different lineage markers that were tested. Remarkably, a high expression of CD123 was observed. To conclude, both CLL-1 and lineage marker expression but not CD123 expression, can be used to detect malignant CD34 þ CD38 cells in follow-up samples from AML patients recovering from chemotherapy. This makes AML stem cell

1704 Table 4 FISH analysis of FACS-sorted CD34+CD38 populations % FISH-positive cells (exact numbers) Cytogenetic analyses Aberrant marker % marker on CD34+CD38 CD34+CD38 marker + CD34+CD38 marker Patient 1 t(8,21) No metaphases a CD19 91 90% (90/100) 0% (0/100) b Patient 2 t(8,21) 100% (10/10 cells) CD56 77 63% (44/70) 13% (3/23) Patient 3 +8 15% (3/20 cells) CD7 85 30% (30/100) 4% (4/100) Abbreviations: FACS, fluorescence-activated cell sorting; FISH, fluorescent in situ hybridization. a Diagnosis of t(8,21) based on routine PCR. b In the CD34+CD38 CD19 fraction the results were corrected for an impurity caused by CD34+CD38+CD19+ cells. On the basis of this and the 90% FISH positivity in the CD34+CD38 CD19+ cells, CD19-negative cells are negative for the translocation. Table 5 Expression of CLL-1, CD123 and lineage markers/marker combinations on normal CD34+CD38 cells NBM median (range, n) RBM median (range, n) CLL-1 0 (0 11, n ¼ 10) 0.0 (0.0 5.0, n ¼ 6) CD123 12 (0 19, n ¼ 6) 60 (50 84, n ¼ 6) CD2 0.1 (0 1.1, n ¼ 4) 3.4 (1.9 7.7, n ¼ 4) CD5 1.1 (0 1.7, n ¼ 4) 4.0 (1.9 7.14, n ¼ 4) CD7 2.0 (0.4 3.6, n ¼ 4) 4.7 (3.6 5.0, n ¼ 4) CD11b 3.4 (2.8 4.9, n ¼ 4) ND CD11c 16 (15;16, n ¼ 2) ND CD19 2.5 (0.3 6.9, n ¼ 4) 2.0 (0.0 5.0, n ¼ 4) CD22 2.1 (0.8 4.8, n ¼ 3) ND CD56 1.7 (0.2 3.6, n ¼ 4) 3.0 (0.0 7.1, n ¼ 3) CD13+CD33 5.8 (0 7.1, n ¼ 3) ND CD13 CD33+ 0 (0 1.9, n ¼ 3) ND HLADR 95 (78 100, n ¼ 4) ND CD13++ 0 (0, n ¼ 3) ND CD33++ 0 (0, n ¼ 3) ND Abbreviations: CLL-1, C-type lectin-like molecule-1; n, number of samples; NBM, normal bone marrow; ND, not done; RBM, regenerating bone marrow. Median, median percentage of marker expression on CD34+CD38 cells. Range, range of percentage marker expression on CD34+CD38 cells., Antigen not expressed; +, antigen expressed; ++, antigen overexpression. MRD detection potentially possible in all patients with CLL-1 expression and in all patients with lineage marker/marker combination expression, excluding CD123; which results in this series in 46/60 CD34-positive AML (77%). Sequential analysis of malignant stem cell populations in AML patients Examples of FACS analysis of malignant CD34 þ CD38 populations after chemotherapy are shown in Figure 3: as expected in RBM marker-negative normal CD34 þ CD38 cells appear at the cost of marker-positive AML CD34 þ CD38 cells in some patients, others show persistence of markerpositive CD34 þ CD38 cells. Figure 4 shows representative examples of follow-up analysis for CLL-1 and lineage marker expression on CD34 þ CD38 cells in combination with whole blast MRD. In these patients with known clinical follow-up, we were able to perform followup analyses with at least three time points after diagnosis, and with whole blast MRD determined in parallel. Not only malignant CD34 þ CD38 cell frequency was analyzed, as a measure for the burden of leukemic CD34 þ CD38 cells but also the fraction of CD34 þ CD38 cells that demonstrated CLL-1 or lineage marker expression. This AML stem cell fraction may represent competition between normal and malignant CD34 þ CD38 cells for the BM niche. The first example (Figure 4a c) shows a patient with a good response to chemotherapy, which was reflected by a gradual reduction in time of both the frequency of leukemic blasts (Figure 4a) and leukemic CD34 þ CD38 cells (Figure 4b) with complete disappearance of aberrant marker expression on the CD34 þ CD38 population (Figure 4c). The second patient achieved complete remission after chemotherapy with low MRD frequency below the threshold of 0.1% (Figure 4d). AML stem cell frequency (Figure 4e) and stem cell fraction (Figure 4f) paralleled whole blast MRD, nevertheless at least 20% of the CD34 þ CD38 cell compartment remained leukemic (Figure 4f). At relapse (time point R) this patient was treated with rescue chemotherapy. Although this resulted in significant blast reduction (Figure 4d), the fraction of malignant CD34 þ CD38 cells continued to increase (Figure 4f), and the patient died of a resistant leukemia. The third patient presented with a low number of blasts at diagnosis, and although complete remission was achieved after allogeneic transplantation, a high number of leukemic blasts remained present (Figure 4g). Stem cell frequency initially dropped (Figure 4h) and later increased, while stem cell fraction showed that almost all CD34 þ CD38 cells were of malignant origin and remained so throughout treatment (Figure 4i). This patient also died of AML. In conclusion, we found that both malignant CD34 þ CD38 cell frequency as well as the malignant fraction of CD34 þ CD38 cells correlated with whole blast MRD and can be used to detect the presence of malignant CD34 þ CD38 cells after chemotherapy, while both types of analysis may even give relevant clinical information, additional to whole blast MRD. Discussion In the present study, we show that in AML and high-risk MDS patients the leukemic CD34 þ CD38 compartment can be detected and quantified in patients in complete remission. Previously, we demonstrated that the frequency of all AML blasts, now referred to as whole blast MRD after achieving complete remission is an important parameter for prognosis in terms of time to relapse and overall survival. Whole blast MRD can be detected using aberrant immunophenotypes, determined on the leukemic cells at diagnosis. We also pointed out that the frequency of CD34 þ CD38 cells at diagnosis correlated with MRD frequency and survival. It was then hypothesized that

1705 Figure 3 Examples of marker defined malignant stem cell populations in paired diagnosis and remission samples. After labelling of the cells with the appropriate antibody combinations, the CD34 þ CD38 cells were identified by gating on the CD45dim/SSClow population and CD34 þ expression using a Boolean gating strategy; subsequently, CLL-1, CD7, CD56 and CD11b expressions were determined in this population. The CD34 þ CD38 population is depicted in the left and right lower quadrants, and the numbers represent the number of events analyzed. (a) CLL-1 expression at diagnosis and (e) CLL-1 expression from the same patient after chemotherapy; in this patient the malignant CLL-1-positive CD34 þ CD38 cells disappeared. (b and f) CD7 expression at diagnosis and after chemotherapy are shown; in this patient the malignant CD7- positive CD34 þ CD38 cells persisted. CD56 expression at diagnosis from another patient is depicted in (c) and after chemotherapy in (g); the malignant CD56 þ CD34 þ CD38 cells have disappeared. (d and h) CD11b expression at diagnosis and in complete remission; CD34 þ CD38 CD11b cells clearly persist. APC, allophycocyanin; CLL-1, C-type lectin-like molecule-1; PE, phycoerythrin. within the MRD compartment, a leukemic stem cell fraction should be present, which is responsible for outgrowth of the leukemic cell population towards relapse. We now proved that within the CD34 þ CD38 stem cell population in BM of AML patients, the leukemic CD34 þ CD38 cells could be specifically detected both at diagnosis and in complete remission using aberrant markers/marker combinations and the new leukemic stem cell marker, CLL-1. In this study, the aberrant markers/marker combinations that are used for whole blast MRD, are shown to be present on the AML CD34 þ CD38 stem cell compartment too, but are virtually absent on NBM CD34 þ CD38 cells. These results might fit in the hypothesis proposed Cozzio et al. 14 and Krivtsov et al. 15 that more differentiated cells, which do not normally have stem cell characteristics, acquire these characteristics upon in vitro introduction of specific genes. Whether this also occurs in vivo remains to be determined. However, this could be an explanation for the presence of aberrant markers, that are normally expressed only on differentiated cells, on leukemic CD34 þ CD38 cells. In comparison with whole blast MRD one main advantage of using lineage markers on CD34 þ CD38 cells to quantify residual disease, is the almost complete lack of expression of these markers in NBM and RBM. Also, the monoclonal antibody CLL-1, owes its value to the fact that it is not expressed on NBM as well as RBM CD34 þ CD38 cells. It could indeed be shown in three patients, using FISH analysis, that cells missing the aberrant marker are likely of normal origin. CD123 seems not very useful in stem cell MRD detection due to the high expression of CD123 in control samples, in steady state but especially in RBM. Whether CD123 expression is still high in long-term follow-up samples has not been determined in this study. The AML stem cell parameters, CD34 þ CD38 frequency and AML CD34 þ CD38 fraction, provided additional information on the quality of remission in a small number of patients studied. Firstly, the CD34 þ CD38 frequency is a representation of the number of malignant CD34 þ CD38 cells that reside in the BM. This number likely reflects the chance of relapse as does whole blast MRD, expressed as frequency within the white blood cell (WBC) compartment and a strong prognostic factor for survival. Secondly, the malignant CD34 þ CD38 fraction offers information about the ratio between normal and malignant cells in the BM and, importantly, might not be affected by the cellularity in the BM. In this study, 44/60 AML patients (73%) showed 450% expression of CLL-1 and/or a lineage marker/marker combination on CD34 þ CD38 cells. A drawback is the absence of a detectable CD34 þ CD38 population in some patients (n ¼ 4). The observation that in two cases without usable aberrancies on the whole blast population (and thus without whole blast MRD detection possible) CLL-1 was expressed on the CD34 þ CD38 population, further indicates the additional value of stem cell analysis. In this respect, it should be mentioned that CLL-1 expression on the whole blast compartment cannot be used for whole blast MRD, since CLL-1 is present on part of the normal CD34 þ CD38 þ progenitor compartment. 11 The cutoff of 50% expression for CLL-1 does not exclude the possibility to discriminate cells with low CLL-1 expression from cells without expression; with homogeneous but low expression of CLL-1 the CD34 þ CD38 population can often be discriminated from the normal CD34 þ CD38 cells. Finally, for AML samples without a detectable CD34 þ population an alternative stem cell population, that is the side population (SP) may be of importance. SP cells have been described to have stem cell characteristics and to give rise to leukemia after transplantation in NOD/SCID mice. 16 In 6/9 AML CD34-negative samples analyzed thus far, SP cells were present and were shown to have lineage marker and/or CLL-1 expression in all six cases, without expression of these markers in 10 NBM controls. The same strategies using CLL-1, lineage marker and CD123 expression might be performed when trying to identify malignant SP cells in follow-up samples of AML patients with such a SP present at diagnosis. In MRD studies, the aberrant marker combinations always contain a myeloid marker. This approach is instrumental to diminish the background staining in NBM and RBM controls. 17

1706 Figure 4 Stem cell parameters and MRD in different AML patients with a different course of the disease. Cells were analyzed at diagnosis and at different time points during follow-up of three AML patients. Patient 1 is depicted in (a c), patient 2 in (d f) and patient 3 in (g i). In (a, d and g) whole blast MRD expressed as a percentage of the total WBC count, is shown. The solid line represents MRD frequency that separates good prognosis patients (o0.1%) from poor prognosis patients (40.1%) according to Feller et al. 1 The dotted line indicates 5% blasts, the cutoff for morphologically defined complete remission. (b, e and h) The frequency of malignant CD34 þ CD38 cells as percentage of the CD45dim population, using a lineage marker or CLL-1; the numbers represent the number of events analyzed. (c, f and i) Malignant CD34 þ CD38 cells as a fraction of the total CD34 þ CD38 population using CLL-1 or aberrant marker expression. The solid line represents the background staining in RBM for the markers that were used (which is zero in (f)). D is diagnosis; R is relapse; 1st is after first course of chemotherapy; 2nd is after second course of chemotherapy; allo is after allogeneic stem cell transplantation. Patient 1 obtained complete remission after one course of chemotherapy. Also the frequency of leukemic blasts (a), the frequency of malignant CD34 þ CD38 cells (b) and the fraction of malignant CD34 þ CD38 cells determined using CD56 expression (c) decreased after chemotherapy. Unfortunately, this patient died of therapy-related complications after the second course of chemotherapy. CLL-1 expression on CD34 þ CD38 cells showed the same pattern (data not shown). Patient 2 suffered from a relapse after the second course of chemotherapy. This was not predicted by conventional MRD measurements (using the 0.1% cutoff) as can be seen in (d). The frequency of malignant CD34 þ CD38 cells within the CD34 þ CD38 population was very low before relapse (b), but the stem cell fraction showed continuous presence of malignant CD34 þ CD38 cells (c). CD56 expression on CD34 þ CD38 cells showed the same results (data not shown). Patient 3 obtained complete remission after the first cycle of chemotherapy, but MRD frequency never reached 0.1%. This is indicative of a fast relapse, which indeed occurred 8 months after the allogeneic stem cell transplantation. MRD frequency remained relatively high(g) and, most remarkable, at all time points the CD34 þ CD38 cells were mainly of malignant origin defined using CD5 expression (h). Unfortunately, at diagnosis, CD5 expression was not determined on the CD34 þ CD38 cells. CLL-1 expression on CD34 þ CD38 cells showed the same results (data not shown). However in the present study, using the NBM and RBM samples, we found that omitting the myeloid markers resulted in very low to zero expression on CD34 þ CD38 cells for all single markers tested. This makes the addition of myeloid markers to the combinations tested on the CD34 þ CD38 cells redundant. For marker combinations that include asynchronous antigen expression (like CD13 þ CD33 or CD13 CD33 þ ) logically the myeloid marker is part of the aberrancy. Since the CD34 þ CD38 population of present interest is so well defined (CD34 þ /CD38 /CD45dim/SSClow/aberrant marker(s)), detection of stem cell MRD requires much less extensive experience than detection of whole blast MRD, which is a laborious job even for experienced researchers and technicians with thorough knowledge of NBM differentiation patterns. 17 The correlation of stem cell MRD with survival needs to be elucidated in a study that includes more patients, and preferably with whole blast MRD determined in parallel. In such a study also the question regarding the number of patients in which it is possible to detect a malignant CD34 þ CD38 population after chemotherapy can be answered. In this study, we were not able to show such results as follow-up analysis was not performed routinely in our patient cohort. In conclusion, we have introduced leukemic stem cell detection in AML and high-risk MDS at diagnosis and in complete remission. Leukemic stem cell isolation at diagnosis and maybe more importantly after chemotherapy can result in its further characterization and might reveal new targets for treatment. To this end assays should be available that are effective with small cell numbers such as flow cytometry to study protein expression, 18,19 and RT-multiplex ligation-dependent probe amplification (RT-MLPA) assays to study gene expression. 20 Lastly, stem cell MRD might be used in future clinical studies to choose optimal time points for therapeutic intervention in patients in remission.

Acknowledgements We acknowledge the Department of Cardiac Surgery and their patients for providing normal bone marrow samples. References 1 Feller N, van der Pol MA, van Stijn A, Weijers GW, Westra AH, Evertse BW et al. MRD parameters using immunophenotypic detection methods are highly reliable in predicting survival in acute myeloid leukaemia. 2004; 18: 1380 1390. 2 Venditti A, Buccisano F, Del Poeta G, Maurillo L, Tamburini A, Cox C et al. Level of minimal residual disease after consolidation therapy predicts outcome in acute myeloid leukemia. Blood 2000; 96: 3948 3952. 3 San Miguel JF, Vidriales MB, Lopez-Berges C, Diaz-Mediavilla J, Gutierrez N, Canizo C et al. Early immunophenotypical evaluation of minimal residual disease in acute myeloid leukemia identifies different patient risk groups and may contribute to postinduction treatment stratification. Blood 2001; 98: 1746 1751. 4 Kern W, Voskova D, Schoch C, Hiddemann W, Schnittger S, Haferlach T. Determination of relapse risk based on assessment of minimal residual disease during complete remission by multiparameter flow cytometry in unselected patients with acute myeloid leukemia. Blood 2004; 104: 3078 3085. 5 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. 6 Costello R, Mallet F, Chambost H, Sainty D, Arnoulet C, Gastaut JA et al. The immunophenotype of minimally differentiated acute myeloid leukemia (AML-M0): reduced immunogenicity and high frequency of CD34+/CD38- leukemic progenitors. 1999; 13: 1513 1518. 7 van Rhenen A, Feller N, Kelder A, Westra AH, Rombouts E, Zweegman S et al. High stem cell frequency in acute myeloid leukemia at diagnosis predicts high minimal residual disease and poor survival. Clin Cancer Res 2005; 11: 6520 6527. 8 Jordan CT, Guzman ML, Noble M. Cancer stem cells. N Engl J Med 2006; 355: 1253 1261. 9 Jordan CT, Upchurch D, Szilvassy SJ, Guzman ML, Howard DS, Pettigrew AL et al. The interleukin-3 receptor alpha chain is a unique marker for human acute myelogenous leukemia stem cells. 2000; 14: 1777 1784. 10 Lowenberg B, Downing JR, Burnett A. Acute myeloid leukemia. N Engl J Med 1999; 341: 1051 1062. 11 Bakker AB, Van den Oudenrijn S, Bakker AQ, Feller N, van Meijer M, Bia JA et al. C-type lectin-like molecule-1: a novel myeloid cell surface marker associated with acute myeloid leukemia. Cancer Res 2004; 64: 8443 8450. 12 Lacombe F, Durrieu F, Briais A, Dumain P, Belloc F, Bascans E et al. Flow cytometry CD45 gating for immunophenotyping of acute myeloid leukemia. 1997; 11: 1878 1886. 13 van der Pol MA, Feller N, Roseboom M, Moshaver B, Westra G, Broxterman HJ et al. Assessment of the normal or leukemic nature of CD34+ cells in acute myeloid leukemia with low percentages of CD34 cells. Haematologica 2003; 88: 983 993. 14 Cozzio A, Passegue E, Ayton PM, Karsunky H, Cleary ML, Weissman IL. Similar MLL-associated leukemias arising from selfrenewing stem cells and short-lived myeloid progenitors. Genes Dev 2003; 17: 3029 3035. 15 Krivtsov AV, Twomey D, Feng Z, Stubbs MC, Wang Y, Faber J et al. Transformation from committed progenitor to leukaemia stem cell initiated by MLL-AF9. Nature 2006; 442: 818 822. 16 Wulf GG, Wang RY, Kuehnle I, Weidner D, Marini F, Brenner MK et al. A leukemic stem cell with intrinsic drug efflux capacity in acute myeloid leukemia. Blood 2001; 98: 1166 1173. 17 Terstappen LW, Loken MR. Myeloid cell differentiation in normal bone marrow and acute myeloid leukemia assessed by multi-dimensional flow cytometry. Anal Cell Pathol 1990; 2: 229 240. 18 van der Pol MA, Broxterman HJ, Pater JM, Feller N, van der MM, Weijers GW et al. Function of the ABC transporters, P-glycoprotein, multidrug resistance protein and breast cancer resistance protein, in minimal residual disease in acute myeloid leukemia. Haematologica 2003; 88: 134 147. 19 van Stijn A, Kok A, van Stalborch MA, van der Pol MA, Feller N, Westra AH et al. Minimal residual disease cells in AML patients have an apoptosis-sensitive protein profile. 2004; 18: 875 877. 20 Hess CJ, Denkers F, Ossenkoppele GJ, Waisfisz Q, McElgunn CJ, Eldering E et al. Gene expression profiling of minimal residual disease in acute myeloid leukaemia by novel multiplex-pcr-based method. 2004; 18: 1981 1988. 1707