The Hematopoietic Stem Cell in Chronic Phase CML Is Characterized by A

Home , Nuclear receptor

The hematopoietic stem cell in chronic phase CML is characterized by a transcriptional proﬁle resembling normal myeloid progenitor cells and reﬂecting loss of quiescence

I Bruns1, A Czibere1, JC Fischer2, F Roels3, R-P Cadeddu1, S Buest1, D Bruennert1, AN Huenerlituerkoglu4, NH Stoecklein5, R Singh1, LF Zerbini6,MJa¨ger7, G Kobbe1, N Gattermann1, R Kronenwett1, B Brors3 and R Haas1

1Department of Hematology, Oncology and Clinical Immunology, Heinrich-Heine-University Du¨sseldorf, Duesseldorf, Germany; 2Institute for Transplantation Diagnostics and Cell Therapeutics, Heinrich-Heine-University Du¨sseldorf, Duesseldorf, Germany; 3Theoretical Bioinformatics, German Cancer Research Center, Heidelberg, Germany; 4Sta¨dtische Kliniken Neuss, Lukaskrankenhaus, Neuss, Germany; 5Department of General and Visceral Surgery, Heinrich-Heine-University Du¨sseldorf, Duesseldorf, Germany; 6BIDMC Genomics Center, Harvard Institutes of Medicine, Boston, MA, USA and 7Department of Orthopedic Surgery, Heinrich-Heine-University Du¨sseldorf, Duesseldorf, Germany

We found that composition of cell subsets within the CD34 þ technology.4 The gene expression profile of the total CD34 þ cell population is markedly altered in chronic phase (CP) cell population was suggestive for an altered composition of the chronic myeloid leukemia (CML). Specifically, proportions and þ þ absolute cell counts of common myeloid progenitors (CMP) and CD34 compartment and flow cytometry-based CD34 cell megakaryocyte–erythrocyte progenitors (MEP) are significantly subset analysis showed a greater proportion of megakaryocyte– greater in comparison to normal bone marrow whereas erythrocyte progenitors (MEP) whereas HSC and GMP subsets absolute numbers of hematopoietic stem cells (HSC) are equal. were proportionally decreased. These observations prompted us To understand the basis for this, we performed gene expres- to perform a more detailed analysis of the molecular signature of sion profiling (Affymetrix HU-133A 2.0) of the distinct CD34 þ CP CML using defined subsets rather than the total population of cell subsets from six patients with CP CML and five healthy þ donors. Euclidean distance analysis revealed a remarkable CD34 stem and progenitor cells. By gene expression profiling transcriptional similarity between the CML patients’ HSC and of CD34 þ subsets from the bone marrow (BM) of six patients normal progenitors, especially CMP. CP CML HSC were with CP CML and five healthy donors, we shed some light on the transcriptionally more similar to their progeny than normal transcriptional deregulation of the leukemic differentiation HSC to theirs, suggesting a more mature phenotype. Hence, the program. In addition, we provide evidence that the CML HSC greatest differences between CP CML patients and normal has a transcriptionally more mature phenotype than its healthy donors were apparent in HSC including downregulation of genes encoding adhesion molecules, transcription factors, counterpart, showing impaired adhesive and migratory capa- regulators of stem-cell fate and inhibitors of cell proliferation cities as well as decreased expression of prominent transcription in CP CML. Impaired adhesive and migratory capacities were factors and stem-cell regulators. Thus, we hypothesize that loss functionally corroborated by fibronectin detachment analysis of HSC quiescence leads to expansion of mature progenitors in and transwell assays, respectively. Based on our findings we CP CML. propose a loss of quiescence of the CML HSC on detachment from the niche leading to expansion of myeloid progenitors. Leukemia (2009) 23, 892–899; doi:10.1038/leu.2008.392; Patients and methods published online 22 January 2009 Keywords: CML; BCR-ABL; stem and progenitor cells; gene expression Patients and cells Informed consent was obtained according to a protocol approved by the ethics committee of Heinrich Heine University from all patients and donors included into this study. We Introduction obtained BM mononuclear cells from six patients with newly diagnosed Ph þ CP CML and five healthy BM donors. The two It has become widely accepted that activation of the BCR-ABL groups did not differ significantly with regard to age, gender and tyrosine kinase is causative for chronic myeloid leukemia body weight. BM samples were processed within 2 h following 1 puncture. CD34 þ cells were positively selected after density (CML). Still, involvement of additional molecular events in 5 the pathogenesis of CML has been demonstrated.2 For instance, gradient centrifugation as previously described. in blast crisis (BC) CML elevated levels of b-catenin lead to expansion of the granulocyte–macrophage progenitor (GMP) Immunofluorescence-based cell sorting subset,2 and inactivation of the transcription factor JunB is able LinÀ, CD34 þ , CD38À HSC and LinÀ, CD34 þ , CD38 þ to increase the number of long-term hematopoietic stem cells lo myeloid progenitor subsets including IL-3Ra CD45RAÀ (LT-HSC) and GMP in a murine model of myeloproliferative lo common myeloid progenitors (CMP), IL-3Ra CD45RA þ disease.3 Previously, we examined CD34 þ stem and progenitor GMP and IL-3RaÀ, CD45RAÀ MEP were isolated from normal cells from patients with chronic phase (CP) CML by microarray and CP CML CD34 þ BM cells by fluorescence-activated cell sorting as previously described.6 Before the samples were Correspondence: Dr I Bruns, Department for Hematology, Oncology subjected to gene expression analysis, we performed fluore- and Clinical Immunology, Heinrich-Heine-University Du¨sseldorf, scence in situ hybridization of the BCR-ABL fusion and Moorenstrae 5, 40255 Duesseldorf, Germany. E-mail: [email protected] quantitatively analyzed the BCR-ABL transcripts by reverse Received 29 July 2008; revised 13 November 2008; accepted 4 transcription (RT)–PCR (details of these analyses are presented in December 2008; published online 22 January 2009 the Supplementary Figure 2). Gene expression profiles of chronic phase CML CD34 þ cell subsets I Bruns et al 893 RNA isolation, amplification and hybridization to turing were done using custom scripts in the PERL scripting microarrays language. A portion of 6 ng of total RNA was used for linear amplification with the Two Cycle Target Labeling Kit (Affymetrix UK Ltd, High Distance Analysis. Unsupervised clustering was performed Wycombe, UK) from highly purified LinÀ, CD34 þ , CD38À on the geometric sample means of the respective CD34 þ cell and LinÀ, CD34 þ , CD38 þ cell subsets. Then 10 mg cRNA subset (HSC, CMP, GMP, MEP) array sets using complete were biotin-labeled using the Enzo BioArray HighYield RNA linkage hierarchical clustering with the Euclidean distance Transcript Labeling Kit (Affymetrix Ltd). Quality control of RNA between CD34 þ cell subset array sets as distance metric. and cRNA was performed using a bioanalyzer (Agilent 2001 Three clusterings were done, one using the combined set of Biosizing; Agilent Technologies, Waldbronn, Germany). Fol- healthy and CML cell subset arrays and two for the separate sets lowing fragmentation, labeled cRNA of each individual patient (healthy and CML, respectively) and the results were presented sample was hybridized to Affymetrix U133 2.0 GeneChips in a tree diagram. covering 14 500 genes and stained according to the manufac- turer’s instructions. Our gene expression analyses fulfilled the recently published consensus guidelines from three European Quantitative real-time reverse transcription polymerase 7 leukemia networks. Array data have been stored in the gene chain reaction expression omnibus database (www.ncbi.nlm.nih.gov/geo/; Corroboration of RNA expression data was performed by real- accession no.: GSE11889) according to MIAME standards. time RT-PCR using the Light Cycler 1.2 (Roche, Mannheim, Germany) as described before.4 GAPDH mRNA served as control for relative quantification. Relative gene expression Quantification, normalization and statistical analysis of levels are presented as the difference of C values of the target gene expression data T gene and GAPDH (DDCT). A list of genes that were confirmed Analysis of differences between distinct subsets. Array with qRT-PCR together with the respective primer sequences is data were analyzed using dChip.8 For normalization, the provided in the Supplementary Table 1. smoothing spline algorithm was performed and expression values were calculated by applying the perfect match–mismatch 8 difference model algorithm. We considered a gene differen- In vitro adhesion and migration analysis tially expressed if the transcript level as a multiple of the control For adhesion analysis, purified LinÀ, CD34 þ , CD38À HSC was 41.5 or o1.5, the lower confidence bound (LCB) 41.2 or (0.5 Â 104) were plated on fibronectin-coated dishes (BD o1.2 and the P value o0.1. The LCB is a stringent estimate of Biosciences, Heidelberg, Germany) in fully supplemented the change in expression as a multiple of the control and has Iscove’s medium as described.14 Cultures were incubated for been shown to be a superior ranking statistic.9 The use of LCB 3 h at 37 1C under 5% CO2 in a humidified atmosphere. provides a 90% confidence that the actual change in expression Nonadherent cells were removed by five washes with phos- as a multiple of the control is a value above the reported LCB. To phate-buffered saline, after which adherent cells were quantified check differential expression of genes identified by linear by light microscopy. amplification and array analysis, we analyzed six genes which Chemotaxis assays were performed using a transwell with a were differentially expressed in normal versus CP CML HSC by pore size of 5 mm (Corning-Costar Corp., Corning Incorporated quantitative RT-PCR (qRT-PCR). We confirmed the results of the Life Sciences, Lowell, MA, USA). Purified LinÀ, CD34 þ , array analysis in all cases and found that the extent of differential CD38À cells (5 Â 104) in fully supplemented Iscove’s medium expression measured by qRT-PCR was in five out of six cases were added to the upper well. Chemotaxis toward 100 ng/ml greater than measured by the microarrays (Supplementary murine stromal-derived factor 1 (SDF-1; R&D Systems, Wiesba- Figure 2). den-Nordenstadt, Germany) in the lower chamber was allowed to continue for 4 h at 37 1C under 5% CO2 in a humidified Pathway analysis. For pathway analysis, data were normal- atmosphere. Cells were harvested from the lower well and ized using the variance stabilizing method10 and summarized stained with a fluorescein isothiocyanate-conjugated (FITC) using median polish. To identify differential expression between monoclonal anti-CD34 antibody (clone 8G12, BD Biosciences, CD34 þ cell subsets, the rank product test was used.11 The rank Heidelberg, Germany). For standardized cell counting 20 000 product test is a nonparametric test, which is particularly calibration beads were added to the cell suspension. The cell applicable for small sample sizes and deviation from assumed suspension was analyzed using a flow cytometer until 10 000 underlying probability distributions in microarray analyses.12 calibration beads were counted. The number of CD34 þ cells Genes were considered to be regulated between two cell subsets was then multiplied by factor 2 and shows the absolute number if the rank product showed a false-positive rate below 0.05. of migrated CD34 þ cells. Based on the number of cells added Genes above this threshold were considered to be not regulated. to the assay the proportion of migrated cells was determined. Genes that were indicated as significantly regulated by the rank Student’s t-test was used to assess significant differences product test were subsequently used to extract pathways from regarding adhesion and cell migration between CML patients the Biocarta database available online at the NCI website (http:// and healthy donors. pid.nci.nih.gov/browse_pathways.shtml#biocarta). To determine the relevance of the resulting pathways the Globaltest13 was performed. Significance was determined by using Cochran’s Q Results and discussion test.13 The significance value associated with the Globaltest results is a measure for how well the group of genes as a whole Common myeloid and megakaryocyte–erythrocyte can distinguish between the conditions under study. Statistical progenitors are expanded in the bone marrow of CP and numerical analysis was performed in the R statistical CML programming environment using the Bioconductor toolset In CML samples, the proportion of primitive HSC within the (http://www.bioconductor.org). Database queries and restruc- CD34 þ stem and progenitor cell population was significantly

Leukemia Gene expression profiles of chronic phase CML CD34 þ cell subsets I Bruns et al 894 smaller (P ¼ 0.0003) with a mean of 2.7% (s.d. 1.8) compared to patients and normal individuals, separately. Then, we compared 12.1% (mean; s.d. 3.5) in the samples obtained from normal the distinct CML subsets with their normal counterparts, to get a donors. Still, in terms of absolute cell numbers, CML and normal better insight into the transcriptional differences in HSC, CMP, HSC were similar with 56.6/ml (mean; s.d. 36.6) and 54.7/ml GMP and MEP. Results of these analyses are detailed below. (mean; s.d. 9.2), respectively (P ¼ 0.95). The proportion of the CMP subset was greater in the samples obtained from CML patients with 26.1% (mean; s.d. 4.6) than in normal donors with 19.7% (mean; s.d. 3.0; P ¼ 0.03), and absolute cell counts in the Lineage differentiation shows similarities of the CML CML samples were 442.9/ml (mean; s.d. 280.9) consistent with a HSC with more mature progenitors 4.7-fold higher amount (P ¼ 0.02). The proportion of GMP in the Looking at the differentiation process we found 178 and 701 BM of patients and normal donors was at almost equal levels genes differentially regulated from HSC to CMP in CML patients with 32.6% (mean; s.d. 7.6) in CML and 34.85% (mean; s.d. 4.7; and normal donors, respectively. The differences observed P ¼ 0.59) in normal donors, whereas a 3.5-fold greater amount between CML patients and normal donors during differentiation m from CMP to MEP or GMP were modest, reflecting the lack of a of 571.8/ l (mean; s.d. 244.8) was found in the CML patients 15 compared to 163.8/ml (mean; s.d. 41.6) in normal donors differentiation block in CP CML. Analyzing the overlap of without reaching statistical significance. The greatest difference genes regulated during transition from HSC to CMP in CML between patients and normal individuals was observed within patients and normal donors, we found only six genes the MEP with a significantly greater proportion of 35.1% (mean; concordantly regulated (Supplementary Figure 3). A pathway s.d. 11.2) in the CML samples than in those from normal donors analysis of the differentially expressed genes showed a (14.94% mean; s.d. 7.0). This difference translated into a 7.7- significant regulation of cell differentiation related genes during fold greater concentration of 549.9/ml in patients with CML transition from HSC to CMP only in normal donors (27/572 (mean; s.d. 193.8; P ¼ 0.009; Figures 1a and b). The data reflect annotated genes, P ¼ 0.006283) and not in CML patients. In a an expansion of myeloid progenitors in CP CML emerging from crosswise analysis we compared the HSC subset of patients with a HSC population with a quantity alike in normal donors. CML and the CMP subset of normal donors and found 325 differentially expressed genes. These were 56% fewer genes than observed between the HSC and CMP subsets in normal hematopoiesis. Although differences between the distinct Gene expression analysis of CML CD34 þ cell subsets subsets seemed to be generally moderate with regard to the Based on the gene expression profiles of the distinct CD34 þ absolute numbers of differentially expressed genes, our findings subsets in six patients with CML and five normal donors, we suggest a particular similarity of the transcriptomes of the CML looked at the transcriptional regulation of the differentiation HSC and normal CMP (Figure 2a). To further corroborate this processes from HSC along the line of differentiation in CML observation, we performed Euclidian distance analysis of the

Healthy CML 163.8/µl 103 103 95.1/µl GMP 2 2 571.8/µl 10 10 54.7/µl 7.0% 0.9% 101 101 HSC CMP 55.0% 59.8% 56.6/µl 71.2/µl 100 100 CD34-PerCP-Cy5 CD34-PerCP-Cy5 HSC Progenitors HSC Progenitors 442.9/µl MEP 549.9/µl 100 101 102 103 100 101 102 103 CD38-FITC

103 103 34.9% 3% 0% 19.7% GMP 2 2 10 10 32.6% 12.1% 29% 43% -PE -PE 1 1 α α 10 10 HSC CMP CMP GMP CMP GMP IL3R IL3R 56% 16% 100 100 2.72% 15.0% MEP MEP 10% 35% 26.1% MEP 35.1% 0 1 2 3 0 1 2 3 Healthy 10 10 10 10 10 10 10 10 CML CD45RA-ECD

Figure 1 (a) Different patterns of hematopoietic stem and progenitor cells in the bone marrow of healthy donors (left) and patients with chronic myeloid leukemia (CML) in chronic phase (right). Immunomagnetic selection of CD34 þ cells followed by multicolor ﬂow cytometry was utilized to analyze hematopoietic stem and progenitor cell subsets. After gating on viable cells and lineage-depletion subfractions of hematopoietic stem cells (HSC; LinÀ, CD34 þ , CD38À), common myeloid progenitors (CMP; LinÀ, CD34 þ , CD38 þ , IL-3Ralo, CD45RAÀ), granulocyte– macrophage progenitor (GMP; LinÀ, CD34 þ , CD38 þ , IL-3Ralo, CD45RA þ ) and megakaryocyte–erythrocyte progenitors (MEP; LinÀ, CD34 þ , CD38 þ , IL-3RaÀ, CD45RAÀ) were determined. Percentages related to LinÀ cells are indicated. Dot plots of one representative experiment out of ﬁve and six in healthy donors and patients, respectively, are shown. (b) Circles represent the mean proportions (top) and absolute cell counts (bottom) of CD34 þ cell subsets of healthy donors (green) and patients (orange) obtained by all experiments. The area of the circles is proportional to the values of the subsets indicated and was normalized to the normal HSC.

Leukemia Gene expression proﬁles of chronic phase CML CD34 þ cell subsets I Bruns et al 895 HSC (Healthy) vs. CMP (Healthy) HSC (CML) vs. CMP (CML) HSC (CML) vs. CMP (Healthy)

11 11 11 10 57 genes 10 10 27 genes 57 genes 9 9 9 8 8 8 7 7 7 6 6 6 5 5 5 4 4 4 644 genes 151 genes 3 3 3 268 genes 2 2 2 1 1 1 0 0 0 Total of 701 differentially expressed genes Total of 178 differentially expressed genes Total of 325 differentially expressed genes

Distance between normal CD34+ cell subsets Distance between CML CD34+ cell subsets Distance between all CD34+ cell subsets

60 30

55 29 60

50 28 HSC(H) HSC(H) 50 45 27 Height Height 40 Height 26 40

35 25 30 30 24

25 23 20 CMP(H) CMP(H) MEP(H) GMP(H) CMP(CML) GMP(CML) HSC(CML) MEP(CML) HSC(CML) CMP(CML) MEP(CML) GMP(CML) GMP(H) MEP(H)

Figure 2 (a) M-A plots showing the number of up- and downregulated genes between hematopoietic stem cells (HSC) and common myeloid progenitors (CMP) of healthy donors (left), HSC and CMP of patients with chronic myeloid leukemia (CML; middle) and HSC of patients with CML and CMP of healthy donors. (b) Tree diagram of CD34 þ cell subsets of healthy donors (left) and healthy donors and CML patients together (right). Clustering was performed on the geometric sample means of the respective CD34 þ cell subsets using complete linkage hierarchical clustering with the Euclidean distance between celltypes as distance metric. whole unselected datasets. This analysis confirmed our assump- L-selectin, CD44 and N-cadherin (CDH2) was significantly tion, as a close proximity of all CML subsets became apparent lower as compared to normal donors. This is in line with (Figure 2b). Notably, looking at the CD34 þ cell subsets from previous reports showing that the BCR-ABL oncogene caused CML patients and healthy donors together, all subsets from downregulation of L-selectin and CXCR4 in hematopoietic patients, including the HSC, cluster around the CMP of healthy cells.18–20 We also found decreased expression levels of genes donors, suggesting a more mature transcriptional phenotype of encoding for members of the CXCR4 downstream signaling HSC in CML patients. In contrast, for normal individuals, the pathway in CML HSC including phosphoinositol-3-kinase cluster tree clearly resembled the model for hierarchical (PI3K)21,22 and 3-phosphoinositide-dependent protein kinase-1 development from the HSC to mature cells with the HSC sitting (PDK1)22 as well as atypical protein kinase C (aPKC).22–24 With at the top of the hierarchy and the more differentiated MEP and regard to cell migration and chemotaxis, genes encoding for GMP subsets at the bottom as proposed by Manz et al.6 All hematopoietic cell-specific DOCK2,25 PDE4,26 and PTP4A127 together, the transcriptome of CML patients and healthy were downregulated in CML HSC (Table 1). These results were individuals apparently differed mainly at the level of HSC and strengthened by a BioCarta-based pathway analysis showing not at later progenitor stages providing further evidence for the differential regulation of the pathways cell to cell adhesion HSC origin of CML and its addiction to oncogenic events signaling, CXCR4 signaling pathway and role of PI3K subunit stimulated by BCR-ABL.16,17 We determined these differences in p85 in actin organization and cell migration in the HSC more detail and found a total of 614 genes differentially of CML patients compared to healthy donors (Figure 3). To expressed between CML patients’ and normal donors’ HSC functionally corroborate these findings, we assessed including 362 genes with a lower expression level and 252 adhesion and transmigration using in vitro assays and found a genes with a higher one, respectively (Supplementary Table 2). significantly reduced proportion of adhesive and migrating HSC Based on these observations and the detailed molecular profile in CML patients compared to their normal counterparts. The of the CML HSC, we suggest in the following a scheme of relative adhesion of CML HSC compared to HSC from healthy malignant transformation in CP CML. donors to the fibronectin-coated dish was reduced to 65.8% (mean, range 50.5–81.7% P ¼ 0.02; Figure 4a). Regarding the migratory capacities we found that 4.1% (mean; range The HSC subset of patients with CML shows impaired 1.4–9.6%) of the HSC subset from patients with CML vs 14.8% migratory and adhesive properties (mean; range 6.6–22.2%; P ¼ 0.01) of the respective cells from In the HSC of patients with CML the gene expression level of the healthy donors migrated through the transwell chamber chemokine receptor CXCR4 and the adhesion molecules (Figure 4b).

Leukemia Gene expression proﬁles of chronic phase CML CD34 þ cell subsets I Bruns et al 896 Table 1 Selected differentially expressed genes in HSC (CML/healthy)

Probe set ID Gene name FC LB FC P value

Adhesion 203440_at CDH2: cadherin 2, type 1, N-cadherin (neuronal) À9.69 À3.81 0.020 211919_s_at CXCR4: chemokine (C-X-C motif) receptor 4* À6.64 À2.28 0.086 204563_at SELL: selectin L (lymphocyte adhesion molecule 1) À2.51 À1.51 0.011 211506_s_at IL8: interleukin 8 3.44 1.57 0.060 204490_s_at CD44: CD44 molecule (Indian blood group) À2.33 À1.31 0.041 208405_s_at CD164: CD164 molecule, sialomucin À2.32 À1.35 0.052 201952_at ALCAM: activated leukocyte cell adhesion molecule À2 À1.24 0.047

Motility 213160_at DOCK2: dedicator of cytokinesis 2 À2.2 À1.54 0.006 203708_at PDE4B: phosphodiesterase 4B À10.49 À4.76 0.023 200732_s_at PTP4A1: protein tyrosine phosphatase type IVA, member 1 À2.55 À1.56 0.029 212363_x_at ACTG1: actin, gamma 1 À1.79 À1.24 0.022 208736_at ARPC3: actin related protein 2/3 complex, subunit 3, 21 kDa À2.1 À1.25 0.055

Transcription 204872_at TLE4: transducin-like enhancer of split 4 À4.62 À2.33 0.022 209959_at NR4A3: nuclear receptor subfamily 4, group A, member 3* À29.75 À2.19 0.160 204621_s_at NR4A2: nuclear receptor subfamily 4, group A, member 2 À6.36 À2.69 0.059 204139_x_at MZF1: myeloid zinc ﬁnger 1 À6.41 À1.98 0.071 222103_at ATF1: activating transcription factor 1 À3.16 À1.84 0.012 204849_at TCFL5: transcription factor-like 5 (basic helix-loop-helix) À3.68 À1.78 0.050 209189_at FOS: v-fos FBJ murine osteosarcoma viral oncogene homolog* À4.45 À1.73 0.067 218486_at KLF11: Kruppel-like factor 11 À3.39 À1.54 0.089 202704_at TOB1: transducer of ERBB2, 1 À2.94 À1.4 0.089 219371_s_at KLF2: Kruppel-like factor 2 (lung) À3.64 À1.29 0.080 214521_at HES2: hairy and enhancer of split 2 (Drosophila) 2.81 1.42 0.032

Stem cell fate 205471_s_at DACH1: dachshund homolog 1 (Drosophila) À2.55 À1.38 0.047 206674_at FLT3: fms-related tyrosine kinase 3 À6.89 À3.3 0.002 204054_at PTEN: phosphatase and tensin homolog À2.38 À1.29 0.075 213792_s_at INSR: insulin receptor À3.45 À1.93 0.016 204304_s_at PROM1: prominin 1* À4.6 À2.4 0.021 203416_at CD53: CD53 molecule À1.81 À1.32 0.010 218968_s_at ZFP64: zinc ﬁnger protein 64 homolog (mouse) À3.33 À1.43 0.088

Cell cycle 210254_at MS4A3: membrane-spanning 4-domains, subfamily A, member 3 À4.62 À1.39 0.990 218740_s_at CDK5RAP3: CDK5 regulatory subunit associated protein 3 2.34 1.28 0.047 215942_s_at GTSE1: G-2 and S-phase expressed 1 4.23 1.37 0.099 206221_at RASA3: RAS p21 protein activator 3 À4.03 À1.52 0.077 Abbreviations: FC, fold change calculated from the mean expression values of six CML HSC versus five normal HSC; LB FC, lower bound of the 90% confidence interval of the mean FC. Genes were considered differentially expressed with a LB FC o1.2, a FC o1.5 and Po0.1. Genes were considered differentially expressed with a LB FC 41.2, a FC 41.5 and Po0.1. *Differential expression confirmed by RT–PCR (Supplementary Figure 2).

CXCR4 signalling Cell to cell adhesion signalling Actin organization and cell migration 2000 Higher expression in CML HSC 200 Higher expression in CML HSC Higher expression in CML HSC Lower expression in CML HSC Lower expression in CML HSC 1500 300 Lower expression in CML HSC p = 0.017 150 p = 0.047 p = 0.006 1000 200 100 Influence Influence Influence 100 5000 50

0 0 0 RHO FAK VCL PAK1 RAS PXN CSK SRC ARP2 ARP3 RAC1 WASP CD31 GNB1 CDC42 ARPC2 ARPC4 ARPC3 PDGFA PIK3R1 GNAQ PDGFR SHP-2 PIK3CA GNAI1 ACTA1 PLCG1 ARPC1B ARPC1A CXCR4 GNGT1 PIK3R1 PIK3CA CXCL12 CTNNB1 CTNNA1 PIK3C2G

Figure 3 Pathway analysis. Green bars show genes with a lower expression in chronic myeloid leukemia (CML) hematopoietic stem cells (HSC) and red bars show genes with a higher expression in CML HSC as compared to the respective normal HSC controls. P values are the result of the Globaltest as described in the Patients and methods section.

Leukemia Gene expression profiles of chronic phase CML CD34 þ cell subsets I Bruns et al 897 1.2 recently been identified as potential upstream regulators of c-Jun * p=0.02 and JunB in a murine AML model,32 we performed qRT-PCR to 1.0 examine the expression levels of NR4A1 and to confirm downregulation of NR4A3. We detected a significant down- 0.8 * regulation of NR4A1 and NR4A3 in HSC of CML patients by 0.6 RT–PCR (Supplementary Figure 2). Of note, NR4A1 and NR4A3 0.4 knockout mice showed expansion of myeloid progenitors, but not HSC resembling more the hematopoietic phenotype of CML

Relative adhesion 0.2 than AML. This observation is in accordance with our quantitative subsets analysis in CP CML patients. Overall, the 0.0 decreased gene expression of three NR4A family members Healthy CML suggests an important role in stem-cell quiescence. In line, we also found decreased transcriptional activity of molecules 25 involved in the maintenance of stem-cell fate such as FLT3 and PTEN (Table 1). Inhibition of FLT3 led to a reduction of the 33 * stem-cell pool and inactivation of PTEN caused long-term 20 decline of HSC fostering differentiation and proliferation34 in murine models. Accordingly, gene expression levels of CD133 were lower in CML HSC compared their normal counterparts. p=0.01 15 Low expression of CD133 glycoprotein is associated with a more mature, proliferative cell fraction.35 We also found a decreased expression of the genes encoding for L-selectin and CD53 in the HSC subset of CML patients (Table 1). These 10

Migration (%) molecules seem to segregate asymmetrically in HSC according * to the model of asymmetric cell division and characterize daughter cells that maintain stem-cell fate.36 Hence, lower 5 expression suggests perturbance of the hematopoietic homeo- stasis that might lead to a greater proportion of differentiating cells than actually needed to replenish the peripheral blood. In 0 summary, these transcriptional changes suggest a loss of Healthy CML quiescence in the CML HSC leading to an increased proliferative activity and expansion of the myeloid progenitors. In accor- Figure 4 In vitro adhesion and transmigration assays. (a) Adhesion of dance with this, we found decreased expression levels of genes hematopoietic stem cells (HSC) to plates coated with ﬁbronectin (FN) associated with the inhibition of cell-cycle progression (Table 1). is shown. The relative adhesion of chronic myeloid leukemia (CML) Among them was member 3 of the membrane-spanning HSC compared to that of HSC from healthy donors is shown. (b) In 37 vitro transmigration of (HSC) from CML patients and healthy donors 4-domains subfamily A, that is able to suppress S-phase entry, toward 100 ng/ml stromal-derived factor 1 (SDF-1). Black and gray as well as Ran/TC4 that causes cell-cycle arrest in G2-phase38 bars represent HSC from patients with CML and healthy donors, and prolongs the duration of the S-phase.39 respectively. Asterisks indicate P values. Considering our ﬁndings and the reexpression of adhesion molecules as CXCR4 in CML stem and progenitor cells following tyrosine kinase inhibitor (TKI) therapy,18,40,41 an eradication of the Ph þ clone could be based on therapeutic principles We therefore hypothesize that impaired adhesive and antagonizing receptor–ligand interactions such as CXCR4 and migratory capacities lead to detachment from and hinder SDF-1. This may be achieved administering the CXCR4 29,42 migration to the stem-cell niche. Thus, deprived of their natural antagonist Plerixafor (AMD 3100) or by enzymatic cleavage habitat, HSC lose quiescence on disruption of the receptor following granulocyte colony-stimulating factor (G-CSF) 43,44 interaction with ligands provided by the niche-building tissues therapy in combination with TKI. Furthermore, the pheno- as shown for CXCR4 and SDF-1.28,29 Interestingly, binding of type of the CML HSC revealed by our analysis suggests a high SDF-1 to CXCR4 increases the activity of the transcription susceptibility to cytotoxic chemotherapy. This might explain the factors JunB30 and c-Fos,31 which we both found downregulated achievement of long-term remissions after high-dose therapy 45 in HSC of CML patients. Therefore, loss of stem-cell quiescence followed by syngeneic BM transplantation in CP CML. on disruption of the ligand–receptor interaction might be mediated by downregulation of transcription factors. Conclusion

The HSC subset of patients with CP CML shows lower Taken together, our findings help to explain that the myeloid gene expression levels of transcription factors and genes CD34 þ subsets expand in CP CML whereas the cell counts of promoting stem-cell self-renewal the more primitive HSC subset equal those of normal donors. Accordingly, we observed a reduced transcriptional activity of The BCR-ABL driven expansion of the myeloid subsets appears transcription factors and transcriptional regulators in the HSC to originate from the HSC, which is unable to enter quiescence. subset of CML patients (Table 1). Among them were the This results in a continuous replenishment of the progenitor activating protein-1 family member c-fos and the nuclear pools. As the progenitors in CP CML do not have a differentia- receptor subfamily 4A members 2 and 3 (NR4A2 and A3), tion block, as reflected by the transcriptional similarity of CML kruppel-like factor family members 2 and 11, C/EBP-b and progenitors and their normal counterparts, their expansion leads myeloid zink finger 1 (MZF1). As NR4A3 and NR4A1 have to an accumulation of maturating white blood cells. The

Leukemia Gene expression profiles of chronic phase CML CD34 þ cell subsets I Bruns et al 898 changes in the HSC, including downregulation of genes 16 Sharma SV, Gajowniczek P, Way IP, Lee DY, Jiang J, Yuza Y et al. involved in the regulation of adhesion and cell motility, stem- A common signaling cascade may underlie ‘addiction’ to the Src, cell self-renewal, differentiation and inhibition of proliferation BCR-ABL, and EGF receptor oncogenes. Cancer Cell 2006; 10: are associated with a transcriptional phenotype resembling that 425–435. 17 Geay JF, Buet D, Zhang Y, Foudi A, Jarrier P, Berthebaud M et al. of normal CMP without losing the immunophenotype of HSC. p210BCR-ABL inhibits SDF-1 chemotactic response via alteration This could also be of relevance for future therapies as TKI of CXCR4 signaling and down-regulation of CXCR4 expression. treatment leads to a reconstitution of adhesion and quiescence Cancer Res 2005; 65: 2676–2683. in the CML stem cell. Eradication of the Ph þ clone could 18 Jongen-Lavrencic M, Salesse S, Delwel R, Verfaillie CM. BCR/ABL- therefore be based on therapeutic principles antagonizing mediated downregulation of genes implicated in cell adhesion and receptor–ligand interactions such as CXCR4 and SDF-1, that motility leads to impaired migration toward CCR7 ligands CCL19 and CCL21 in primary BCR/ABL-positive cells. Leukemia 2005; 19: is, by blockage of CXCR4 or enzymatic cleavage following 373–380. administration of G-CSF in combination with TKI. 19 Ptasznik A, Urbanowska E, Chinta S, Costa MA, Katz BA, Stanislaus MA et al. Crosstalk between BCR/ABL oncoprotein Acknowledgements and CXCR4 signaling through a Src family kinase in human leukemia cells. J Exp Med 2002; 196: 667–678. The following sources of financial support to IB and AC are greatly 20 Wang JF, Park IW, Groopman JE. Stromal cell-derived factor- 1alpha stimulates tyrosine phosphorylation of multiple focal appreciated: Leuka¨mie Liga e.V. Du¨sseldorf, Forschungskommis- adhesion proteins and induces migration of hematopoietic sion der Medizinischen Fakulta¨tDu¨sseldorf. We thank Anne Koch progenitor cells: roles of phosphoinositide-3 kinase and protein and Sabrina Pechtel for excellent technical assistance and Dr kinase C. Blood 2000; 95: 2505–2513. Bernd Giebel and Dr Frank Neumann for fruitful discussions. 21 Ganju RK, Brubaker SA, Meyer J, Dutt P, Yang Y, Qin S et al. The alpha-chemokine, stromal cell-derived factor-1alpha, binds to the References transmembrane G-protein-coupled CXCR-4 receptor and activates multiple signal transduction pathways. J Biol Chem 1998; 273: 1 Savona M, Talpaz M. Getting to the stem of chronic myeloid 23169–23175. leukaemia. Nat Rev Cancer 2008; 8: 341–350. 22 Le Good JA, Ziegler WH, Parekh DB, Alessi DR, Cohen P, Parker 2 Jamieson CH, Ailles LE, Dylla SJ, Muijtjens M, Jones C, Zehnder JL PJ. Protein kinase C isotypes controlled by phosphoinositide et al. Granulocyte-macrophage progenitors as candidate leukemic 3-kinase through the protein kinase PDK1. Science 1998; 281: stem cells in blast-crisis CML. N Engl J Med 2004; 351: 657–667. 2042–2045. 3 Passegue E, Wagner EF, Weissman IL. JunB deficiency leads to a 23 Petit I, Goichberg P, Spiegel A, Peled A, Brodie C, Seger R et al. myeloproliferative disorder arising from hematopoietic stem cells. Atypical PKC-zeta regulates SDF-1-mediated migration and Cell 2004; 119: 431–443. development of human CD34+ progenitor cells. J Clin Invest 4 Diaz-Blanco E, Bruns I, Neumann F, Fischer JC, Graef T, Rosskopf 2005; 115: 168–176. M et al. Molecular signature of CD34(+) hematopoietic stem and 24 Nakanishi H, Brewer KA, Exton JH. Activation of the zeta isozyme progenitor cells of patients with CML in chronic phase. Leukemia of protein kinase C by phosphatidylinositol 3,4,5-trisphosphate. 2007; 21: 494–504. J Biol Chem 1993; 268: 13–16. 5 Czibere A, Prall WC, Zerbini LF, Jager M, Kobbe G, Knipp S et al. 25 Fukui Y, Hashimoto O, Sanui T, Oono T, Koga H, Abe M Exisulind induces apoptosis in advanced myelodysplastic syn- et al. Haematopoietic cell-specific CDM family protein DOCK2 drome (MDS) and acute myeloid leukaemia/MDS. Br J Haematol is essential for lymphocyte migration. Nature 2001; 412: 2006; 135: 355–357. 826–831. 6 Manz MG, Miyamoto T, Akashi K, Weissman IL. Prospective 26 Kohyama T, Liu X, Wen FQ, Zhu YK, Wang H, Kim HJ et al. PDE4 isolation of human clonogenic common myeloid progenitors. Proc inhibitors attenuate fibroblast chemotaxis and contraction of native Natl Acad Sci USA 2002; 99: 11872–11877. collagen gels. Am J Respir Cell Mol Biol 2002; 26: 694–701. 7 Staal FJ, Cario G, Cazzaniga G, Haferlach T, Heuser M, Hofmann 27 Fiordalisi JJ, Keller PJ, Cox AD. PRL tyrosine phosphatases regulate WK et al. Consensus guidelines for microarray gene expression rho family GTPases to promote invasion and motility. Cancer Res analyses in leukemia from three European leukemia networks. 2006; 66: 3153–3161. Leukemia 2006; 20: 1385–1392. 28 Cashman J, Clark-Lewis I, Eaves A, Eaves C. Stromal-derived factor 8 Li C, Wong WH. Model-based analysis of oligonucleotide arrays: 1 inhibits the cycling of very primitive human hematopoietic cells expression index computation and outlier detection. Proc Natl in vitro and in NOD/SCID mice. Blood 2002; 99: 792–799. Acad Sci USA 2001; 98: 31–36. 29 Nie Y, Han YC, Zou YR. CXCR4 is required for the quiescence of 9 Li C, Hung Wong W. Model-based analysis of oligonucleotide primitive hematopoietic cells. J Exp Med 2008; 205: 777–783. arrays: model validation, design issues and standard error 30 Zhu W, Boachie-Adjei O, Rawlins BA, Frenkel B, Boskey AL, application. Genome Biol 2001; 2: RESEARCH0032. Ivashkiv LB et al. A novel regulatory role for stromal-derived 10 Huber W, von Heydebreck A, Sultmann H, Poustka A, Vingron M. factor-1 signaling in bone morphogenic protein-2 osteogenic Variance stabilization applied to microarray data calibration and differentiation of mesenchymal C2C12 cells. J Biol Chem 2007; to the quantification of differential expression. Bioinformatics 282: 18676–18685. 2002; 18 (Suppl 1): S96–S104. 31 Joo EK, Broxmeyer HE, Kwon HJ, Kang HB, Kim JS, Lim JS et al. 11 Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank products: a Enhancement of cell survival by stromal cell-derived factor-1/ simple, yet powerful, new method to detect differentially regulated CXCL12 involves activation of CREB and induction of Mcl-1 and genes in replicated microarray experiments. FEBS Lett 2004; 573: c-Fos in factor-dependent human cell line MO7e. Stem Cells Dev 83–92. 2004; 13: 563–570. 12 Saeys Y, Inza I, Larranaga P. A review of feature selection 32 Mullican SE, Zhang S, Konopleva M, Ruvolo V, Andreeff M, techniques in bioinformatics. Bioinformatics 2007; 23: Milbrandt J et al. Abrogation of nuclear receptors Nr4a3 and 2507–2517. Nr4a1 leads to development of acute myeloid leukemia. Nat Med 13 Goeman JJ, van de Geer SA, de Kort F, van Houwelingen HC. A 2007; 13: 730–735. global test for groups of genes: testing association with a clinical 33 Weisel KC, Yildirim S, Schweikle E, Kanz L, Mohle R. Effect of outcome. Bioinformatics 2004; 20: 93–99. FLT3 inhibition on normal hematopoietic progenitor cells. Ann NY 14 Adams GB, Chabner KT, Alley IR, Olson DP, Szczepiorkowski ZM, Acad Sci 2007; 1106: 190–196. Poznansky MC et al. Stem cell engraftment at the endosteal niche 34 Zhang J, Grindley JC, Yin T, Jayasinghe S, He XC, Ross JT et al. is specified by the calcium-sensing receptor. Nature 2006; 439: PTEN maintains haematopoietic stem cells and acts in lineage 599–603. choice and leukaemia prevention. Nature 2006; 441: 518–522. 15 Kavalerchik E, Goff D, Jamieson CH. Chronic myeloid leukemia 35 Toren A, Bielorai B, Jacob-Hirsch J, Fisher T, Kreiser D, Moran O stem cells. J Clin Oncol 2008; 26: 2911–2915. et al. CD133-positive hematopoietic stem cell ‘stemness’ genes

Leukemia Gene expression proﬁles of chronic phase CML CD34 þ cell subsets I Bruns et al 899 contain many genes mutated or abnormally expressed in leukemia. 41 Jin L, Tabe Y, Konoplev S, Xu Y, Leysath CE, Lu H et al. CXCR4 up- Stem Cells 2005; 23: 1142–1153. regulation by imatinib induces chronic myelogenous leukemia 36 Beckmann J, Scheitza S, Wernet P, Fischer JC, Giebel B. (CML) cell migration to bone marrow stroma and promotes Asymmetric cell division within the human hematopoietic stem survival of quiescent CML cells. Mol Cancer Ther 2008; 7: 48–58. and progenitor cell compartment: identiﬁcation of asymmetrically 42 Liles WC, Broxmeyer HE, Rodger E, Wood B, Hu¨bel K, Cooper S segregating proteins. Blood 2007; 109: 5494–5501. et al. Mobilization of hematopoietic progenitor cells in healthy 37 Donato JL, Ko J, Kutok JL, Cheng T, Shirakawa T, Mao XQ et al. volunteers by AMD3100, a CXCR4 antagonist. Blood 2003; 102: Human HTm4 is a hematopoietic cell cycle regulator. J Clin Invest 2728–2730. 2002; 109: 51–58. 43 Nervi B, Link DC, DiPersio JF. Cytokines and hematopoietic stem 38 Ren M, Coutavas E, D’Eustachio P, Rush MG. Effects of mutant cell mobilization. J Cell Biochem 2006; 99: 690–705. Ran/TC4 proteins on cell cycle progression. Mol Cell Biol 1994; 44 Hattori K, Heissig B, Wu Y, Dias S, Tejada R, Ferris B et al. 14: 4216–4224. Placental growth factor reconstitutes hematopoiesis by recruiting 39 Milano Jr J, Strayer DS. Effects of overexpression of VEGFR1(+) stem cells from bone-marrow microenvironment. Nat Ran/TC4 mammalian cells in vitro. Exp Cell Res 1998; 239: Med 2002; 8: 841–849. 31–39. 45 Kanfer EJ, Buckner CD, Fefer A, Storb R, Appelbaum FR, Hill RS 40 Fruehauf S, Topaly J, Schad M, Paschka P, Gschaidmeier H, Zeller et al. Allogeneic and syngeneic marrow transplantation following WJ et al. Imatinib restores expression of CD62L in BCR-ABL- high dose dimethylbusulfan, cyclophosphamide and total body positive cells. J Leukoc Biol 2003; 73: 600–603. irradiation. Bone Marrow Transplant 1987; 1: 339–346.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia