Expression of HOX genes, HOX cofactors, and MLL in phenotypically ...

Leukemia (1999) 13, 687–698  1999 Stockton Press All rights reserved 0887-6924/99 $12.00 http://www.stockton-press.co.uk/leu

Expression of HOX genes, HOX cofactors, and MLL in phenotypically and functionally defined subpopulations of leukemic and normal human hematopoietic cells H Kawagoe, RK Humphries, A Blair, HJ Sutherland and DE Hogge Terry Fox Laboratory, British Columbia Cancer Agency, and Department of Medicine, University of British Columbia, Vancouver, BC, Canada

To explore the possibility that deregulated HOX gene expression might commonly occur during leukemic hematopoiesis, we compared the relative levels of expression of these and related genes in phenotypically and functionally defined subpopulations of AML blasts and normal hematopoietic cells. Initially, a semi-quantitative RT-PCR technique was used to amplify total cDNA from total leukemic blast cell populations from 20 AML patients and light density cells from four normal bone marrows. Expression of HOX genes (A9, A10, B3 and B4), MEIS1 and MLL was easily detected in the majority of AML samples with the exception of two samples from patients with AML subtype M3 (which expressed only MLL). Low levels of HOXA9 and A10 but not B3 or B4 were seen in normal marrow while MLL was easily detected. PBX1a was difficult to detect in any AML sample but was seen in three of four normal marrows. Cells from nine AML patients and five normal bone marrows were FACS-sorted into CD34ⴙCD38ⴚ, CD34ⴙCD38ⴙ and CD34ⴚ subpopulations, analyzed for their functional properties in long-term culture (LTC) and colony assays, and for gene expression using RT-PCR. 93 ⴞ 14% of AML LTC-initiating cells, 92 ± 14% AML colony-forming cells, and ⬎99% of normal LTC-IC and CFC were CD34ⴙ. The relative level of expression of the four HOX genes in amplified cDNA from CD34ⴚ as compared to CD34ⴙCD38ⴚ normal cells was reduced ⬎10-fold. However, in AML samples this down-regulation in HOX expression in CD34ⴚ as compared to CD34ⴙCD38ⴚ cells was not seen (P ⬍ 0.05 for comparison between AML and normal). A similar difference between normal and AML subpopulations was seen when the relative levels of expression of MEIS1, and to a lesser extent MLL, were compared in CD34ⴙ and CD34ⴚ cells (P ⬍ 0.05). In contrast, while some evidence of down-regulation of PBX1a was found in comparing CD34ⴚ to CD34ⴙ normal cells it was difficult to detect expression of this gene in any subpopulation from most AML samples. Thus, the down-regulation of HOX, MEIS1 and to some extent MLL which occurs with normal hematopoietic differentiation is not seen in AML cells with similar functional and phenotypic properties. Keywords: AML; HOX; MLL; MEIS1; PBX1a; hematopoiesis

Introduction The clustered homeobox or HOX genes encode transcription factors which were first recognized as playing a major role in regulating embryonic cell fate in lower organisms.1,2 In mammals, 39 different HOX genes are organized into four clusters (HOXA, B, C and D) located on four different chromosomes. In embryonic development their expression patterns are spatially and temporally restricted according to their 3⬘ to 5⬘ position on the chromosome.3 A number of studies have suggested a role for HOX genes in normal hematopoietic development.4 For example, the patterns of expression of some members of the HOXA and B gene Correspondence: DE Research Center, 601 Canada; Fax: 604 877 Received 2 December

Hogge, Terry Fox Laboratory, BC Cancer West 10 Avenue, Vancouver, BC V5Z 1L3, 0712 1998; accepted 5 February 1999

clusters in hematopoietic cells are closely related to their stage of differentiation. 5⬘ HOX genes such as HOXA9 and A10 are highly expressed among CD34+ cells, including the subpopulations enriched for both primitive and lineage-committed progenitors, and down-regulated in more differentiated CD34− cells. HOX genes with a more 3⬘ location, such as HOXB3 and B4, are down-regulated within the CD34+ population as cells leave the most primitive cell compartment and express lineage commitment.5 The functional importance of these changes has been examined by engineering altered expression of individual HOX genes in mouse bone marrow cells. Overexpression of HOXA10 increased the megakaryocyte colony-forming ability of mouse marrow cells and decreased the numbers of macrophage and pre-B-lymphoid progenitors while a significant proportion of the mice developed acute myelogenous leukemia (AML) after 5 to 10 months.6 The targeted disruption of HOXA9 resulted in reduction of total leukocyte counts and a blunted response to granulocyte colony-stimulating factor (GCSF).7 In contrast, overexpression of HOXB4 selectively expanded primitive murine progenitors with multilineage marrow repopulating ability without detectable effects on hematopoietic differentiation.8 On the other hand, up-regulation of HOXB3 expression perturbed both T and B lymphoid development and myelopoiesis leading to a myeloproliferative disorder in mice.9 Analysis of samples directly isolated from patients with leukemia has suggested a role for HOX genes and their cofactors in malignant hematopoiesis. In AML associated with the chromosomal rearrangement t(7;11)(p15;p15) HOXA9 is fused to nucleoporin NUP98.10 Transcripts for HOXA9, A10, B3 and B4 are detected in malignant cells from various subtypes of AML, but interestingly, expression of at least one of these (A10) is not seen in acute promyelocytic leukemia.11,12 PBX1a, a DNA binding partner of several HOX gene family members,13 is fused to E2A in acute lymphoblastic leukemia (ALL) containing the t(1;19).14 Another HOX binding partner, Meis1,15 is activated by retroviral insertion together with Hoxa7 and Hoxa9 in myeloid leukemia in BXH-2 mice.16 Finally, MLL, the human homologue of the Drosophila trithorax gene and a positive regulator of HOX gene expression, is expressed in normal human hematopoietic cells and may be required for the proper differentiation of myeloid progenitors.17,18 It is also frequently rearranged in AML and ALL in association with translocations involving chromosome 11q23.19,20 Although the malignant blasts in AML are characterized by deregulated proliferation and blocks at various levels of terminal hematopoietic cell differentiation21 it has been possible to demonstrate that the hierarchical organization of progenitor cell types that is seen in normal hematopoiesis22–25 persists in these diseases.26–29 Thus, both normal and leukemic progenitors will form colonies in methylcellulose, initiate long-term

Hox genes in AML H Kawagoe et al

688

hematopoiesis when placed in coculture with fibroblast feeder layers, and engraft in immunodeficient mice. It has been possible to purify subpopulations of cells from normal and AML sources that are highly enriched for these progenitor cell types based on their expression of cell surface markers including CD34. The more primitive progenitors (including normal longterm culture-initiating cells (LTC-IC), leukemic suspension culture-initiating cells, normal human progenitors which competitively repopulate nonobese diabetic/severe combined immune deficiency (NOD/SCID) mice, and SCID mouse leukemia-initiating cells) are preferentially detected in the CD34+ CD38− subpopulation while the more differentiated colonyforming cells (CFC) are frequently CD34+ CD38+.5,24,25,27,29 Progenitors of all kinds are virtually absent from the CD34− cell fraction in normal cell samples and in most cases of AML.27,30,31 As discussed above, it appears likely that the HOX family of transcription factors is important in the orderly progression of normal events in hematopoietic differentiation. Thus, it seemed relevant to investigate how the expression of these genes in the development of AML progenitors compared with that seen in their normal counterparts. In this study, phenotypically and functionally defined subpopulations of AML and normal cells were isolated and compared for their expression of HOX genes, HOX cofactors and MLL. Although, as expected, the FACS-sorted normal and malignant subpopulations showed similar functional properties in assays for LTCIC and clonogenic progenitors the pattern of expression of all the genes studied differed between normal and leukemic cells. These results add to a growing body of data demonstrating deregulated expression of HOX and related genes in human acute leukemias and suggesting a possible central role for the HOX regulatory pathway in the etiology of these disorders. Materials and methods

Cell samples Normal human bone marrow cells from donors for allogeneic transplantation and samples from patients with newly diagnosed AML (peripheral blood cells or bone marrow cells) were obtained after informed consent and with the approval of the Clinical Research Ethics Board of the University of British Columbia. Low-density cells (⬍1.077 g/cm3) were isolated using a Ficoll/Hypaque gradient. Cells were immediately subjected to analysis as described below or frozen in Iscove’s medium containing 50% fetal calf serum (FCS) and 10% dimethylsulphoxide (DMSO), and stored at −135°C. Morphological, histochemical, and cytogenetic analyses were performed on initial leukemic samples as part of the routine clinical diagnostic procedures.

Cell sorting Cells were suspended in Hank’s medium with 2% FCS and 0.1% sodium azide (HFN) at 5 × 107 cells/ml and incubated on ice for 30 min with fluorescence isothiocyanate (FITC)conjugated monoclonal antibody to CD34 (8G12-FITC; Terry Fox Laboratory, Vancouver, Canada) and phycoerythrin (PE)conjugated monoclonal antibody to CD38 (CD38-PE, Becton Dickinson, San Jose, CA, USA). After washing and suspension in HFN with 1 mg/ml of propidium iodine (PI), cells were sorted on a FACStar plus (Becton Dickinson) into

CD34+CD38−, CD34+CD38+, and CD34− subpopulations based on their fluorescence intensity. Comparison with unstained controls was used to set gates excluding ⬎99% of negatively staining cells after gating for viability based on lack of PI uptake as previously described.27,31 The same gating strategy was used to sort both normal and AML cells. Cells sorted into Iscove’s medium containing 50% FCS were then subjected to further analysis as described below.

cDNA generation and amplification Representative total amplified cDNA was generated from cell samples with an oligo(dT)-based primer and poly(A) tailing strategy as previously described.5 Briefly, 1 × 104 cells were lysed in guanidinium isothicyanate and the nucleic acids precipitated by ethanol. For reverse transcription, the pellet was resuspended in a solution containing 6 ␮l of diethyl pyrocarbonate-treated water, 2 ␮l of 5 × RT buffer (GIBCO/BRL, Gaithersberg, MD, USA), 1 ␮l of 0.1 m dithiothreitol, 0.2 ␮l of 25 mm dNTPs (Pharmacia, Uppsala, Sweden), 0.2 ␮l of a special 60-mer oligo(dT) primer (1 ␮g/␮l),32 0.1 ␮l of RNase inhibitor (10 units/␮l, GIBCO/BRL), and 0.5 ␮l of SuperScript II reverse transcriptase (200 units/␮l, GIBCO/BRL). The sample was incubated at 40°C for 1 h, heated to 70°C for 10 min, and ethanol precipitated. The washed pellet was resuspended in 5.5 ␮l of tailing solution (1 ␮l of 5 × TdT buffer (GIBCO/BRL), 0.5 ␮l of 100 mm dATP (Pharmacia), 3.5 ␮l of water, and 0.5 ␮l terminal deoxynucleotidyl transferase (15 units/␮l, GIBCO/BRL)), and incubated at 37°C for 15 min and then 70°C for 10 min. For cDNA amplification by polymerase chain reaction (PCR) 5 ␮l of the above solution was mixed with 5 ␮l of 10 × Taq buffer (GIBCO/BRL), 5 ␮l of 50 mm MgCl2, 4 ␮l of the 60-mer primer, 0.5 ␮l of nuclease-free bovine serum albumin (BSA) (10 mg/ml, Boehringer Mannheim), 0.25 ␮l of Triton X100, 2 ␮l of d(GCT) deoxynucleotide (each 25 mm), 26.75 ␮l of water, and 1 ␮l of Taq DNA polymerase (5 units/␮l, GIBCO/BRL). PCR reaction was performed with an Ericomp thermal cycler (Ericomp, San Diego, CA, USA) using the following conditions: 94°C for 1 min; 37°C for 2 min; 72°C for 10 min followed by 40 cycles of 94°C for 1 min; 55°C for 2 min; 72°C for 10 min.

Southern blot analysis of amplified cDNA Amplified cDNA mixtures were electrophoresed on a 1% agarose gel, denatured, and then transferred to a nylon membrane. The probe for HOXA9 was generated by PCR from cDNA generated from the HL-60 human leukemic cell line using the following amplimers: 5⬘ primer CTGTTGATGGTAGGCTGTAT (GenBank accession number U41813 nucleotides 753–772) and 3⬘ primer AGGTGGAGAAAATGATGAAT (nucleotides 1120–1111). The probe for HOXB3 was a fragment from a HOXB3 cDNA which was cloned from a cDNA library generated from CD34+ normal hematopoietic cells (GenBank accession number U59298). The probes for HOXA10, HOXB4 and PBX1a were made respectively from a HOXA10 cDNA (kindly provided by Dr C Largman, Department of Veterans Affairs, Medical Center, San Francisco, CA, USA), a HOX B4 cDNA (provided by Dr E Boncinelli, Ospedale S Faffaele, Milan, Italy) and a PBX1a cDNA (provided by Dr K Monica, Department of Pathology, Stanford University, Stanford, CA, USA). All of the HOX gene


probes were free of homeodomain sequences. The probe for PBX1a did not hybridize to PBX2 and three cDNAs (also kindly provided by Dr K Monica) blotted onto nylon membrane under the stringency of conditions used in this study. The MLL probe was a cDNA fragment located in the 3⬘ coding region (kindly provided by Dr NJ Zeleznik-Le, Department of Medicine, University of Chicago, Chicago, IL, USA). A homeodomain-free fragment of the Meis1 cDNA (kindly provided by Dr T Nakamura, Cancer Institute, Tokyo, Japan) was used as a probe for MEIS1. The nucleotide sequence of the cDNA fragment of Meis1 used was ⬎90% identical to that of MEIS1. This probe detected the previously reported expression pattern of Meisl in mouse tissues under the hybridization conditions used in this study.33 The ␤-actin cDNA fragment used in control hybridizations was used as previously described.5 Probes were labeled with 32P-dCTP using a random primers DNA labeling system (GIBCO/BRL) according to the manufacturer’s direction. Blots were hybridized with radiolabeled probes at 60°C overnight in 4.4 × SSC (NaCl 0.3 m, Na3C0H5O7 0.03 m, pH 7), 7.5% formamide, 0.75% SDS, 1.5 mm EDTA, 0.75% skim milk, 370 ␮g/ml denatured salmon sperm DNA, and 7.5% dextran sulfate. After hybridization, blots were washed twice in 0.3 × SSC, 0.1% SDS at 60°C for 30 min. Densitometric analysis were performed using the STORM 860 phosphoimaging system (Molecular Dynamics, Sunnyvale, CA, USA). All densitometric values were normalized against those of ␤-actin from the same samples.

Colony assays FACS-sorted normal or leukemic cells, either initially or after long-term culture (LTC), were plated in methylcellulose medium (0.92% methylcellulose, 30% FCS, 2 mm l-glutamine, 10−4 m ␤-mercaptoethanol, 1% BSA, in ␣-medium; (StemCell Technologies, Vancouver, Canada) with the following human growth factors: 3 U/ml human erythropoietin (Epo; StemCell Technologies), 10 ng/ml granulocyte–macrophage colony-stimulating factor (GM-CSF; Sandoz Pharma, Basel, Switzerland), 10 ng/ml interleukin-3 (IL-3; Sandoz Pharma), 50 ng/ml stem cell factor (SF; Terry Fox Laboratory), and 50 ng/ml flt-3 ligand (FL; Immunex, Seattle, WA, USA). After 14 days of incubation at 37°C in 5% CO2, colonies containing more than 20 cells were counted using inverted microscopy.

Long-term cultures Human marrow feeders (HMF) were obtained as previously described.34 These were irradiated with 15 Gy of 250 kilovolt (peak) X-rays and then seeded into tissue culture dishes at 3 × 104 cells/cm2. Sl/Sl-IL-3 fibroblasts were generated by transduction of the human IL-3 cDNA into mouse Sl/Sl fibroblasts as previously described.35 These cells produce bioactive human IL-3 at a concentration of 16 ng/ml as determined by the ability of their growth medium to stimulate 3H-Tdr incorporation into MO7e cells. Mixed feeders are a 1:1 mixture of M2-10B4 cells, a cloned line of mouse bone marrow origin, engineered to produce human IL-3 (4 ng/ml) and G-CSF (190 ng/ml), and Sl/Sl cells engineered to produce human IL3 (1 ng/ml) and SF (4 ng/ml).35 Sl/Sl-IL-3 and mixed feeders were irradiated to 80 Gy and then plated into tissue culture dishes at 3 × 104 cells/cm2. FACS sorted AML cells were suspended in Myelocult LTC medium (StemCell Technologies) with 10−6 m solucortef (Sigma-Aldrich Canada, Oakville,

Ontario, Canada), and plated on to the HMF or Sl/Sl-IL-3 feeders. HMF were used for LTC of all AML samples except those from patients 10 and 15. In these two latter cases Sl/Sl-IL-3 feeders were used, and LTC medium was supplemented with FL (50 ng/ml) at each weekly half medium change. These LTC conditions for the various AML cell samples were selected based on previous experience with unsorted cells from the same source.28 FACS-sorted normal bone marrow cells were plated on to mixed feeders as described.35 Cultures were maintained at 37°C in 5% CO2 with weekly half medium changes. After 5 weeks, both nonadherent and adherent cells were harvested by trypsinization, and plated into methylcellulose as described above to determine the clonogenic cell content of each LTC.

Fluorescence in situ hybridization (FISH) FACS-sorted AML cells were cytocentrifuged onto slides, and then fixed in 3:1 methanol and acetic acid. Colonies from methylcellulose assay were plucked into 75 mm KCl and fixed on to multiwell glass slides (Celline, New Field, NJ, USA). The DNA probe D8Z2 (ATCC, Rockville, MD, USA) for chromosome 8 was labeled with digoxygenin (DIG: BoehringerMannheim, Mannheim, Germany) using a nick translation kit (GIBCO/BRL). The 11q23 MLL probe labeled with DIG was purchased from Oncor (Gaithersburg, MD, USA). The yeast artificial chromosome (YAC) clone containing 550 kb of human DNA encompassing the break point on 16p13 which occurs in the inv(16) rearrangement (CEPHy904E02854) was obtained from Max-Planck Institute (Berlin, Germany). The human DNA in this YAC clone was amplified by inter-Alu PCR as described previously,36 and then labeled with DIG by nick translation. The slides were pretreated in 2 × SSC at 37°C for 30 min, denatured in 70% formamide in 4 × SSC at 75°C for 2 min, and dehydrated in ethanol. The chromosome 8 centromere probe, at a concentration of 2 ng/␮l in 78% formamide, 14% wt/vol dextran sulphate, 1.4 × SSC, and 100 ␮g/ml salmon sperm DNA, was denatured at 75°C for 5 min and then applied to denatured samples on slides. The inv(16) YAC probe, at a concentration of 20 ng/␮l in 71% formamide, 14% wt/vol dextran sulphate, 3 × SSC, and 40 ␮g/ml human Cot-1 DNA (GIBCO/BRL), was denatured at 75°C for 5 min, and then incubated at 37°C for 30 min to allow annealing of repetitive DNA sequences before application to slides. Hybridization was performed at 37°C in a moist chamber overnight. Slides hybridized with 8 centromere probe were washed once in 50% formamide, 4 × SSC, twice in 55% formamide, 4 × SSC, once in 2 × SSC, and twice in 0.1 × SSC at 45°C for 15 min each. For the inversion 16 YAC probe, slides were washed twice in 55% formamide, 4 × SSC, once in 2 × SSC, and then once in 0.5 × SSC each for 15 min at 45°C. The 11q23 probe from Oncor was hybridized and washed according to the manufacturer’s instructions. For signal detection, slides were incubated with 8 ␮g/ml sheep anti-DIG-FITC antibody (Boehringer-Mannheim) in 1% BSA, 4 × SSC at 37°C for 1 h. Slides were then washed at room temperature (RT) in 4 × SSC, 4 × SSC with 0.1% Triton X-100, and PN buffer (0.1 m NaH2PO4Na2HPO4, pH 8.0, 0.1% Nonidet-P40), each for 10 min. To amplify the signal, slides were incubated with 30 ␮g/ml rabbit anti-sheep-FITC (Vector Labs, Burlingame, CA, USA) in PMN buffer (PN buffer with 5% nonfat dry milk powder) at 37°C for 1 h. Slides were washed three times in PN buffer at RT for 5 min each and then in 4 × SSC with 0.1% Triton X-100 at 40°C for 5 min. Nuclear DNA was

689


690

counterstained with 0.5 ␮g/ml PI in antifade (200 mm 1,4diazabicyclo-[2,2,2]-octane, 2 mm Tris, pH 8, in 90% glycerol). A Zeiss Axioplan fluorescence microscope with double bandpass filters was used for observation and analysis. For cytospin slides of sorted cells, a minimum of 55 cells which had clear signals was analyzed to determine the proportion of abnormal cells. In the case of plucked colonies, a minimum of five cells with a clear signal were analyzed within each colony, and it was classified as normal or abnormal when the same hybridization signal was present in at least 80% of the cells analyzed.

Statistical analysis Comparison of the relative expression levels of each gene between AML and normal cells was performed by Student’s t-test. Statistical significance was assigned when the probability that there was no difference between two variables was ⬍0.05.

After normalization for ␤-actin expression in the same sample, HOX A9 and A10 transcripts were each detected in 14 of 20 AML samples with the highest levels seen among samples with the M5 and M4 subtypes. However, in two of four M4eo samples (Nos 11 and 12) expression of these genes was undetectable and in one the level of expression was very low (No. 13). HOX A9 and A10 could also be detected in three of four normal marrow samples but at lower levels than that typically seen in the AML samples. In general, HOX B3 and B4 expression was more difficult to detect in all samples tested and, in fact, expression of both genes was undetectable in all four normal marrows. However, HOX B3 and B4 expression was detected in 11 and 13 of 20 AML samples, respectively. Similar results were obtained on analysis for MEIS1 with expression undetectable in normal marrows but easily detected in 13 of 20 AML samples. In contrast, PBX1a transcripts could only be detected in three of 20 AML samples but were present in three of four normal marrows while MLL expression was easily detected in 22 of 24 samples including all four normal marrows. Interestingly, the two patient samples with AML FAB subtype M3 failed to show expression of any of the genes studied except MLL.

Results

Expression of HOX and related genes in unsorted AML peripheral blood cells and normal bone marrow

Isolation of FACS-sorted subpopulations of leukemic and normal cells

Using the RT-PCR procedure described in the Methods total cDNA was amplified from light density cells from the peripheral blood or bone marrow of 20 patients with newly diagnosed AML and four normal bone marrow samples (Table 1).

Subpopulations of light density cells from nine AML samples and five normal marrows were isolated based on their expression of CD34 and CD38. The characteristics of the AML patients from whom the samples were taken are shown in

Table 1

Expression of HOX, MEIS1, PBX1a and MLL in unsorted AML and normal hematopoietic cells

Pt. No.

AML 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Normal marrow 1 2 3 4

Ratio to ␤-actina

FAB

M1 M1 M2 M3 M3 M4 M4 M4 M4 M4 M4eo M4eo M4eo M4eo M5 M5 M5 M5 M5 RAEB

HOXA9

HOXA10

HOXB3

HOXB4

MEIS1

PBX1a

MLL

0.1 negb neg neg neg 1 0.5 0.01 2 0.1 neg neg 0.01 0.3 0.02 2 0.2 0.3 0.03 0.09

0.1 neg neg neg neg 0.7 0.1 0.04 1 0.4 neg neg 0.01 0.3 0.02 2 0.4 0.4 0.09 0.4

neg neg neg neg neg 0.2 0.05 0.02 0.5 neg 0.05 neg 0.05 0.01 neg 0.10 0.03 0.08 0.01 neg

0.03 neg neg neg neg 0.1 0.03 0.04 0.1 neg 0.1 neg 0.03 0.08 neg 0.06 0.04 0.04 0.05 0.01

0.05 neg neg neg neg 0.03 neg 0.02 0.02 0.01 0.1 0.03 0.08 0.03 neg 0.2 neg 0.02 0.02 0.03

neg neg neg neg neg neg neg neg neg 0.07 neg neg 0.01 neg neg neg neg neg neg 0.03

0.08 0.02 0.03 0.3 0.02 0.07 0.02 0.05 0.03 0.1 0.2 neg 0.07 0.08 neg 0.1 0.01 0.02 0.09 0.1

0.02 neg 0.01 0.02

0.03 neg 0.1 0.02

neg neg neg neg

neg neg neg neg

neg neg neg neg

0.1 neg 0.07 0.1

0.1 0.2 0.04 0.04

Densitometric value for each gene/densitometric value for ␤-actin. neg, below the limit of detection in the RT-PCR assay estimated at ⬍0.01 of actin level.

a

b


Table 2

691

Patient characteristics

Pt. No.

6 10 11 13 14 15 19 20 21

Age

Sex

Tissue

FAB

WBC (% blast) ×109/l

69 77 59 53 38 75 38 68 65

F M M F M F M M F

PB PB PB PB PB PB BM PB PB

M4 M4 M4 M4eo M4eo M5 M5a RAEB M4eo

65 (83) 56 (51) 76 (⬎90) 18 (35) 59 (79) 260 (⬎90) 68 (93a) 63 (6) 88 (70)

Bone marrow cytogenetics

+8(50/50)b t(12,22),+8,del(20q)(15/15) inv(16)(p13q22)(13/14) inv(16)(p13q22)(25/25) inv(16)(p13q22)(7/15),+22(8/15) del(11)(q14q25)(15/15) rea(11)(q23)(15/15) +8+8(4/30) inv(16)(p13q22)(25/25)

PB, peripheral blood; BM, bone marrow. a Percentage from bone marrow. b Numbers of abnormal/metaphase analyzed.

Table 2. These samples were selected because the malignant cells carried a cytogenetic marker which was detectable by FISH on interphase cells and/or because a proportion of cells in the malignant sample expressed the antigens of interest. Table 3 shows the proportion of light density nucleated cells in each of the subpopulations CD34+CD38−, CD34+CD38+ and CD34−. For the AML samples the mean percentage of total cells with each phenotype was 19%, 36% and 42%, respectively, while for normal bone marrow the corresponding values were 0.62%, 6.3% and 93%, respectively. A large proportion (22–79%) of the cells in each sorted fraction from the eight AML samples where such analysis was possible were cytogenetically abnormal by FISH.

of AML-CFC (Table 4). One patient sample (No. 11) did not form discrete colonies in these assays. From cultures of the remaining eight patient samples a mean ± s.d. of 92 ± 14% of all AML-CFC detected were CD34+. The fold enrichment of AML CFC in the CD34+CD38− fraction as compared to CD34− cells in the same sample varied from 0.3- to ⬎1480-fold while a similar comparison between CD34+CD38+ and CD34− cells demonstrated a 1.3- to ⬎730-fold enrichment for the same progenitor cell type. The majority of AML CFC were cytogenetically abnormal by FISH from every patient sample except No. 20. However, in this patient with refractory anemia with excess blasts (RAEB) only a minority of bone marrow metaphase cells contained the +8+8 abnormality for which the FISH analysis was done. For detection of AML LTC-IC, cells from the same sorted populations were placed in coculture with irradiated feeder layers for 5 weeks followed by assay of the contents of the culture in methylcellulose colony assay as previously described (Table 4).28 No colonies formed in assays from 5week-old LTC of cells from patients 11 and 14. However, from

Functional analysis of subpopulations of leukemic and normal hematopoietic cells A portion of the AML cells from each FACS-sorted fraction were placed directly into methylcellulose assay for detection Table 3

Percentage of total light density cells in AML and normal subpopulations

CD34+CD38−

Sample No.

AML 6 10 11 13 14 15 19b 20 21

CD34+CD38+

CD34−

%

FISHa

%

FISH

%

FISH

6 20 22 29 49 14 1.0 11 23

(41/66) (109/138) (69/100) (46/100) (52/100) (74/100)

63 31 45 18 33 35 68 2 33

(42/76) (88/114) (75/100) (69/100) (43/100) (58/77)

30 48 31 52 16 49 30 85 43

(26/55) (78/100) (65/100) (29/100) (55/100) (73/95)

(55/214) (36/100)

(22/99) (51/100)

Mean ± s.d.

19 ± 14

36 ± 20

42 ± 19

Normal BM 1 2 3 4 5

0.5 0.5 0.7 1.0 0.4

2.8 6.0 7.2 7.9 7.7

96 93 92 91 92

Mean ± s.d.

0.62 ± 0.24

6.3 ± 2.1

93 ± 1.9

a

Numbers of abnormal cells/numbers of cells analyzed. The cytogenetic abnormality in this patient sample could not be detected with available FISH probes.

b

(90/117) (58/100)


692

Table 4

CFC and LTC-IC analysis of sorted cell populations

Progenitors per 105 cellsa (FISH analysis)b CD34+CD38− AML 6 10 13 14 15 19 20 21

Normal BM 1 2 3 4 5

CFC LTC-IC CFC LTC-IC CFC LTC-IC CFC CFC LTC-IC CFC LTC-IC CFC LTC-IC CFC LTC-IC

53 18 181 7 195 2 47 10 68 2230 58 6700 496 1480 1

LTC-IC LTC-IC LTC-IC LTC-IC CFC LTC-IC

79 000 10 600 55 000 110 000 6 000 112 000

(5/9) (2/8) (10/10) (8/9) (8/15) (0/6) (6/11) (4/4) (3/10) (2/10) (0/6) (6/12) (0/8)

CD34+CD38+

151 2 1464 2 151 ⬍0.2 7 13 ⬍1 1019 3 9700 600 730 ⬍0.1 2 000 2 600 600 1 300 21 600 3 600

(7/7) (2/4) (10/10) (4/4) (12/19) (5/5)

(0/10) (0/8) (10/15)

CD34−

6 ⬍0.4 585 ⬍0.2 1 ⬍0.4 ⬍1 ⬍1 ⬍1 782 2 44 21 ⬍1 ⬍0.05

(4/4) (8/9)

(0/10) (0/5)

⬍4 26 ⬍4 ⬍4 6 ⬍1

a

Numbers of CFC per 105 cells placed initially in methylcellulose assay or LTC for 5 weeks prior to assay for CFC content. Numbers of cytogenetically abnormal colonies/numbers of colonies analyzed by FISH.

b

the remaining seven patients’ samples 93 ± 14% of the LTCderived CFC came from the CD34+ cell fractions and in the case of every sample except that from patient No. 20 the large majority of these colonies were from LTC of the CD34+CD38− subpopulation. As compared to CD34− cells from the same sample the fold enrichment of LTC-IC in the CD34+CD38− fraction varied from ⬎5.4 to ⬎68 while that for CD34+CD38+ cells varied from 1.5- to ⬎8.7-fold. With FISH analysis it was possible to confirm the leukemic origin of at least some of the LTC-IC detected from three patient samples. More than 99% of LTC-IC from all five normal marrows were CD34+ and in four more than 96% were CD34+CD38−. In the one normal marrow studied for CFC content the majority of these progenitors were CD34+CD38+ and less than 1% were CD34−, consistent with our previous experience.30

Expression of HOX and related genes in FACS-sorted subpopulations of leukemic and normal cells RT-PCR-generated total cDNA from the AML and normal cell populations studied in functional assays were analyzed for expression of HOX and related genes. Representative examples of Southern blots of normal and AML samples hybridized with probes for these eight genes are shown in Figure 1a and b, respectively. Table 5 shows the expression data for the nine AML patients and five normal marrow samples analyzed. After normalization for ␤-actin expression the densitometry values obtained for CD34+CD38+ and CD34− cells were divided by those obtained for CD34+CD38− cells to produce the values shown in Table 5. Figures 2 and 3 show the mean relative levels of expression of the different genes in normal and leukemic cell populations and the results of statistical comparisons. It is evident from Figures 1a, 2 and Table

5 that expression of HOX A9, A10, B3 and B4 in normal cells is ⭓10-fold higher in the CD34+CD38− subpopulation than in CD34− cells. Expression of HOXB3 and B4 was also significantly lower in the CD34+CD38+ as compared to the CD34+CD38− population but this decline was not seen for HOXA9 and A10. These results are consistent with those previously published which demonstrated down-regulation of specific HOX gene expression with the differentiation of primitive normal hematopoietic cells.5 In contrast, as shown in Figures 1b, 2 and Table 5 such down-regulation generally does not occur in AML cells. Although analysis was attempted for all four HOX genes in all nine patient samples, the levels of HOX A9 expression in samples 11 and 13, the level of HOX B3 in sample 10, and HOX B4 in samples 10 and 15 were too low in all three sorted fractions to allow the comparisons to be made. Nevertheless, for each AML sample the analysis was completed for at least two of the four HOX genes studied. When the mean relative levels of HOX gene expression in CD34− normal and AML cells were compared highly significant differences were apparent for HOX A10 (P = 0.0042) and HOX B3 (P = 0.0038) while less striking, but still significant, differences were observed for HOX A9 (P = 0.021) and HOX B4 (P = 0.039). When relative levels of expression were compared between CD34+CD38+ normal and AML cells significantly higher levels of expression were seen for HOX10 (P = 0.034), HOX B3 (P = 0.025) and HOX B4 (P = 0.042) but not for HOX A9 in the leukemic cells. As shown in Table 5, in individual AML samples decreased expression of specific HOX genes in CD34− as compared to CD34+CD38− cells to within the 95% confidence interval of the mean relative level seen with normal cells was rarely observed. In sample 20 HOX B3 and B4 expression were down-regulated in CD34− cells to relative levels of 0.032 and


693

Figure 1 Southern blot analysis of total amplified cDNA from FACS-sorted AML and normal bone marrow. Cells were sorted into CD34+CD38−, CD34+CD38+ and CD34− fractions and total cDNA was amplified using RT-PCR from 1 × 104 sorted cells as described in Materials and methods. Each sample with or without reverse transcription (RT+ or RT−) was sequentially hybridized with specific probes for each gene and control ␤-actin probe. Representative results from (a) normal bone marrow and (b) AML (patient 2) are shown.

0.058, respectively. However, HOXA9 and A10 expression in the same specimen was not significantly reduced in CD34− cells (relative levels of expression 0.42 and 0.47, respectively). Although HOX B4 expression was relatively reduced in CD34− cells from patient 13 (relative level 0.030), HOX A10 and B3 expression were not (relative expression in CD34− cells 0.12 and 0.58, respectively). Thus, in each of the nine leukemic samples analyzed lack of significant down-regulation of expression of at least one of the four HOX genes analyzed was observed in CD34− as compared to CD34+ CD38− malignant cells and in seven samples this was seen for all of the HOX genes whose expression could be detected.

Expression of HOX regulatory genes and cofactors in sorted cells MEIS1, a DNA binding cofactor for HOXA9 and A10, shows a similar pattern of expression to that of the HOX genes, ie it is reduced with differentiation as normal cells leave the CD34+CD38− compartment but is expressed at consistent levels in all three subpopulations derived from AML cells (Figure 3a). On average, these differences between leukemic and normal cells reach statistical significance for both CD34+CD38+ (P = 0.041) and CD34− cells (P = 0.043) as compared to the CD34+CD38− population. However, in two of nine AML samples (Nos 13 and 14) MEIS1 expression was reduced in CD34− as compared to CD34+CD38− cells to an extent similar

to that which was seen with the normal marrow samples (relative expression levels 0.050 and 0.056, respectively, for each sample). Thus, HOXA9 and A10 expression showed a similar pattern to that of MEIS1 expression (ie lack of significantly reduced expression in CD34− cells) in six of seven and seven of nine AML cases analyzed, respectively. Similar studies were done with another HOX cofactor, PBX1a. Expression of this gene was very difficult to detect in cDNA derived from AML cells, either from the total light density cell population (Table 1) or FACS-sorted subpopulations (Table 5). However, PBX1a transcripts were detected in all five normal marrow samples with the highest levels seen in CD34+CD38− cells and approximately 4- and 9-fold lower levels of expression seen in CD34+CD38+ and CD34− normal cells, respectively (Figure 3b). In the two patient samples (Nos 19 and 20) where PBX1a expression could be detected in the sorted subpopulations, No. 20 gave results similar to those seen in normal marrow (expression in CD34+CD38+ and CD34− cells relative to CD34+CD38− cells 0.14 and 0.05, respectively) while No. 19 showed similar levels of expression in all subpopulations (relative levels of expression in CD34+CD38+ and CD34− as compared with CD34+CD38− cells, 1.6 and 1.5, respectively). When the same type of analysis was done for MLL, a positive regulator of HOX gene expression, this gene was found to be expressed at relatively constant levels in all three sorted subpopulations from four normal and eight leukemic cell samples (Figure 3c). There was, however, a modest reduction

0.12 0.87 0.28 0.092 0.13

0.3 0.33

Normal BM 1 2 3 4 5

Mean ±95% CI

0.026 0.019

0.03 0.062 0.014 0.0064 0.016

1.1 0.93 neg neg 0.29 0.057 2.29 0.42 1.55

0.35 0.18

0.2 0.68 0.39 0.23 0.23

1.1 2.1 1.1 0.43 0.61 0.52 1.43 3.3 0.17

CD34− CD34+CD38+

HoxA10

0.04 0.018

0.068 0.048 0.02 0.04 0.022

0.6 1.1 1.3 0.12 0.27 0.07 0.83 0.47 1.27

0.13 0.12

neg 0.06 0.11 0.31 0.047

3.3 neg 0.89 0.79 0.61 0.52 1.76 0.42 0.64

CD34− CD34+CD38+

HoxB3

0.048 0.047

neg 0.0057 0.06 0.11 0.017

0.41 neg 0.5 0.58 0.58 0.72 1.23 0.032 1.27

0.2 0.11

0.16 0.36 neg 0.1 0.17

3.1 neg 1 0.15 0.7 neg 1.53 1.8 0.19

CD34− CD34+CD38+

HoxB4

0.087 0.061

0.15 0.13 neg 0.045 0.023

1.02 neg 1.07 0.03 0.32 neg 1.93 0.058 0.9

0.15 0.039

0.19 0.097 0.17 0.11 0.19

0.72 2.1 0.95 0.21 0.43 0.48 1.1 3.7 0.08

CD34− CD34+CD38+

Meis1

0.038 0.024

0.08 0.03 0.02 0.05 0.011

0.41 2.2 0.27 0.05 0.056 0.47 0.58 0.22 0.87

0.2 0.082

0.2 0.27 0.19 0.29 0.054

neg neg neg neg neg neg 1.6 0.14 neg

CD34− CD34+CD38+

Pbx1

0.093 0.073

0.16 0.2 0.022 0.063 0.019

neg neg neg neg neg neg 1.5 0.05 neg

0.41 0.33

0.25 0.82 neg 0.23 0.32

1.8 1.44 1 0.14 1.18 neg 1.11 4.1 0.57

0.34 0.14

0.51 0.34 neg 0.23 0.29

2 0.81 1.1 0.11 0.55 neg 1.05 0.27 0.86

CD34− CD34+CD38+ CD34−

MLL

neg, not detected. a Data shown on this table were generated using densitometry to compare the intensity of phosphoimaged signals from Southern blots of total cDNA from each sample. The relative intensity of the signal for the gene of interest from each sorted cell subpopulation was compared to the signal intensity for ␤-actin from the same sample. The value thus obtained for CD34+CD38− from each patient or normal cell sample was compared to the value obtained for CD34+CD38− and CD34− cells from that sample to derive the values shown. Patient values shown in bold are those which fall within the 95% confidence intervals (CI) of the mean value for the normal bone marrows (BM).

3.8 1.34 neg neg 0.53 0.57 3.2 11.5 0.52

CD34+CD38+

HoxA9

Fold-change in expression relative to CD34+CD38− cells

Relative levels of expression of Hox genes and their cofactors in subpopulations of AML and normal hematopoietic cellsa

Patient 6 10 11 13 14 15 19 20 21

Table 5


694


695

Figure 2 Relative expression levels of HOX genes in AML and normal bone marrow subpopulations. Densitometric analysis was carried out on each blot and all data were normalized to those of ␤-actin from the same samples. The resulting values were used to determine the relative expression levels of HOXA9 (a), A10 (b), B3 (c) and B4 (d) in CD34− and CD34+CD38+ cells. The values shown indicate the mean expression levels relative to the level in CD34+CD38− cells.

in MLL expression in CD34− normal cells which was significantly different (P = 0.048) than the expression seen in CD34− AML cells. Discussion The block in differentiation and deregulation of cell growth that characterizes the malignant clone in AML has been recognized for many years. However, the molecular mechanism by which most of the genetic abnormalities identified in AML blasts influence the functional abnormalities in leukemic cells remains unclear. If, as has been suggested, the HOX regulatory pathway is of central importance in normal hematopoietic development disruptions of this pathway, either directly or indirectly, might be a common feature of human leukemias.37 Consistent with that hypothesis, we and others have detected expression of various HOX genes in AML blasts at levels generally higher than those seen in normal bone marrow cells (Table 1).11,12 However, HOX expression is present at low or undetectable levels in acute promyelocytic leukemia where the blasts exhibit more obviously differentiated myeloid features and occasionally in AML samples of other subtypes (Table 1). Furthermore, HOX transcripts can be detected at increased levels in normal hematopoietic cells when subpopulations containing only the most primitive progenitors are analyzed.5 These findings suggested that the apparent differences between leukemic and normal cells could reflect the relatively undifferentiated state of most AML blasts, a higher progenitor cell concentration in malignant as compared to

normal hematopoietic tissues, or unusually high levels of HOX expression in the malignant progenitors themselves. The current series of experiments was undertaken to discriminate between these various alternatives and to extend our analysis of gene expression to the HOX cofactors and regulatory genes. Previous work documented the expression of HOX genes in normal bone marrow cells defined by their expression of CD34 and several antigens which are absent from primitive progenitors such as LTC-IC but are variably expressed on more mature, lineage-committed CFC. These studies demonstrated down-regulation of HOX gene expression that occurred; firstly, and most strikingly, as cells differentiated and left the CD34+ cell compartment and secondly, for genes such as HOXB3 and B4, as the most primitive cells defined by the phenotype CD34+CD38− differentiated to become CD38+. Recently, it has become possible to apply many of the same assays and cell sorting strategies used with normal cells to the analysis of malignant cells from patients with AML. For example, in most AML patient samples the most primitive progenitors which initiate long-term suspension culture or growth in immunodeficient mice are CD34+CD38− while most AMLCFC are also CD34+ but frequently do express CD38.27,29 With these studies on normal and leukemic cells in mind, in the current study we chose to isolate three sorted subpopulations of AML and normal cells defined by their expression of CD34 and CD38 for comparison of relative levels of expression of genes involved in the HOX regulatory pathway. The five normal marrows tested confirmed the results described in the previous report by demonstrating reduced levels of expression for the four HOX genes analyzed among the more differentiated cells (Figures 1 and 2).5


696

Figure 3 Relative expression levels of MEIS1, PBX1a and MLL in AML and normal bone marrow subpopulations. Relative expression levels of MEISI (a) and PBX1a (b), and MLL (c) in CD34− and CD34+CD38+ as compared to CD34+CD38− subpopulations were determined by the same method as described in Figure 2. Since PBX1a expression was not detected in seven of nine AML samples results are shown for normal cells only in (b).

When sorted subpopulations of AML cells were isolated for RNA extraction and cDNA amplification a large proportion of the cells were cytogenetically abnormal, as were most AMLCFC. The finding that many of the CFC derived from AML LTC were normal by FISH analysis is consistent with our previous demonstration of the emergence of cytogenetically normal progenitors in LTC of AML cells.28 However, in each AML sample the large majority of all LTC-IC and 100% of those which were cytogenetically abnormal were CD34+ and most of these were also CD38−. Although a significant minority of AML-CFC from two patient samples were CD34⫺, ⬎95% of these cells in six other specimens were CD34+, consistent with our previous experience.27 In contrast to normal marrow, down-regulation of HOX expression was rarely seen when comparing CD34+CD38− cells, which contain the majority of the most primitive AML progenitors, to either the CD34− population, or CD34+CD38+ cells. Thus, HOX gene expression is deregulated throughout the hierarchy of malignant hematopoiesis and includes AML cells which have differentiated to the point that progenitor activity is difficult or impossible to detect. None of the nine AML samples studied here had evidence for direct involvement of any HOX genes or their known cofactors in genetic rearrangements. However, the HOX genes themselves are infrequent partners in chromosomal translocations.10 Relatively little is known about the factors which regulate HOX expression and even less is understood about the target genes for these transcription factors. Nevertheless, many of the HOX cofactors which have been discovered, such as PBX1a and Meis1, have been implicated in leukemogenic

events. For example, coactivation of Meis1 (the murine homologue of MEIS1) and Hoxa7 and a9 by retroviral insertion has been demonstrated in murine AML.16 As shown in Figures 1, 2 and 3, concordant deregulation of MEIS1 and its known DNA binding partners HOXA9 and A10 is a frequent occurrence in human AML. Other investigators have reached similar conclusions when comparing expression of MEIS1 and HOXA7 in total blast cell populations from patients with AML.38 PBX1a is involved in translocations in human ALL and is a known binding partner for many HOX genes including those studied here.13 Interestingly, the pattern of PBX1a expression in subpopulations of normal cells was similar to that for the four HOX genes and MEIS1. However, its expression was difficult to detect in AML blasts. When transcripts could be detected the level of expression in the various cell subpopulations was not consistently different from that which was observed in normal marrow cells. Recently, the overexpression of Hoxb4 in combination with Pbx1 has been reported to transform Rat-1 fibroblasts more effectively than overexpression of Hoxb4 alone.39 However, given the current results it seems unlikely that this mechanism is frequently important in human myeloid leukemogenesis. MLL, a known positive regulator of HOX expression, is often rearranged in acute leukemias (eg patient 15 in the current series).40 Thus, it was of interest to examine the pattern of expression of this gene in normal and malignant human hematopoietic cells. Consistent with its apparent role in hematopoiesis.18 MLL transcripts were easily detected in normal marrow. However, there were rather modest differences in the


levels of expression either between the different subpopulations of normal cells or between normal marrow and the AML samples. In Mll−/− embryos the most striking defect in yolk sac hematopoiesis is a reduction in the number and size of granulocyte and/or macrophage colonies formed. Thus, Mll may be necessary for the survival and/or proliferation of hematopoietic progenitors but not for their differentiation. This suggests that some HOX-mediated effects on hematopoietic development may be regulated by pathways that do not involve MLL. In any case it appears unlikely from our results that the altered expression of HOX genes in AML can be explained by downstream effects secondary to MLL deregulation. It is already apparent that a variety of factors other than MLL are involved in HOX gene regulation.41,42 For example, retinoic acid profoundly affects HOX expression patterns in embryonic development and expression of several members of the polycomb group of genes which negatively regulate HOX expression is seen in normal hematopoietic cells.43 It seems possible that these or other genes involved in HOX regulation may ultimately be found to be aberrantly expressed or rearranged in AML, initiating signals that converge on HOX genes as downstream targets. Expression of multiple HOX family members would be deregulated by such events which might have more profound effects than disruption of a single HOX gene. Since deregulated expression was seen for at least two of the four HOX genes analyzed and/or MEIS1 for each of the nine AML patients studied here, the current data support such a mechanism. Previous studies have shown that although deregulation of a single HOX gene may cause hematopoietic abnormalities it may be insufficient to cause frank leukemia.7,8 Even mice transplanted with bone marrow cells engineered to overexpress HOXA10 develop AML only after a long latent period suggesting the necessity for secondary genetic changes.6 However, recently it has been shown that although overexpression of Hoxa9 alone or in combination with Pbx1b is not leukemogenic overexpression of both Hoxa9 and Meis1 in mouse hematopoietic cells causes AML in transplanted mice within several months.44 Our failure to detect HOXA10 expression in some unsorted leukemic blast samples with AML subtypes M1, M2 and M4 while others have observed at least low levels of expression of this gene in such specimens11 may reflect the different sensitivity of the techniques used and/or subtle differences in the quality of mRNA from the various cells studied. However, it may also reflect some degree of heterogeneity in HOX gene expression that can only be revealed when larger numbers of samples from different AML subtypes are analyzed. It is also possible that this difference in HOX expression among AML samples with the same FAB subtype may be associated with other specific biological or clinical features of the disease. However, even greater numbers of samples would be required to identify such associations. Finally, in this report only the expression of seven of the many HOX genes and cofactors were analyzed raising the possibility that expression of other related and relevant genes was simply overlooked in the apparently negative samples. In summary, altered expression patterns for four different HOX genes and one of their cofactors, MEIS1, is a common feature of human AML in spite of the heterogeneity of clinical and laboratory features and known genetic rearrangements that characterize these diseases. It seems reasonable to suppose that these aberrantly expressed genes and the products of their target genes may work synergistically to influence the character of the transformed cells. As more is learned about

the normal regulation of HOX expression and the precise nature of the genes whose expression they in turn control, the relative importance of these changes and the cellular processes they perturb should become apparent. If disruption of the normal HOX regulatory pathway is not only a common feature of AML but of fundamental importance to leukemogenesis it may even be possible to design therapeutic strategies that target these abnormalities.

Acknowledgements This work was supported by a grant from the National Cancer Institute of Canada with funds from the Terry Fox Foundation.

References 1 Gehring WJ. Homeoboxes in the study of development. Science 1987; 236: 1245–1252. 2 Kessel M, Gruss P. Murine developmental control genes. Science 1990; 249: 374–379. 3 Graham A, Papalopulu N, Krumlauf R. The murine and drosophila homeobox gene complexes have common features of organization and expression. Cell 1989; 57: 367–378. 4 Lawrence HJ, Largman C. Homeobox genes in normal hematopoiesis and leukemia. Blood 1992; 80: 2445–2453. 5 Sauvageau G, Lansdorp PM, Eaves CJ, Hogge DE, Dragowska WH, Reid DS, Largman C, Lawrence HJ, Humphries RK. Differential expression of homeobox genes in functionally distinct CD34+ subpopulations of human bone marrow cells. Proc Natl Acad Sci USA 1994; 91: 12223–12227. 6 Thorsteinsdottir U, Sauvageau G, Hough MR, Dragowska W, Lansdorp PM, Lawrence HJ, Largman C, Humphries RK. Overexpression of HOXA10 in murine hematopoietic cells perturbs both myeloid and lymphoid differentiation and leads to acute myeloid leukemia. Mol Cell Biol 1997; 17: 495–505. 7 Lawrence HJ, Helgason CD, Sauvageau G, Fong S, Izon DJ, Humphries RK, Largman C. Mice bearing a targeted interruption of the homeobox gene HOXA9 have defects in myeloid, erythroid and lymphoid hematopoiesis. Blood 1997; 89: 1922–1930. 8 Sauvageau G, Thorsteinsdottir U, Eaves CJ, Lawrence HJ, Largman C, Lansdorp PM, Humphries RK. Overexpression of HOXB4 in hematopoietic cells causes the selective expansion of more primitive populations in vitro and in vivo. Genes Dev 1995; 9: 1753–1765. 9 Sauvageau G, Thorsteinsdottir U, Hough MR, Hugo P, Lawrence HJ, Largman C, Humphries RK. Overexpression of HOXB3 in hematopoietic cells causes defective lymphoid development and progressive myeloproliferation. Immunity 1997; 6: 13–22. 10 Borrow J, Shearman AM, Stanton VPJ, Becher R, Collins T, Williams AJ, Dube I, Katz F, Kwong YL, Morris C et al. The t(7;11)(p15;p15) translocation in acute myeloid leukaemia fuses the genes for nucleoporin NUP98 and class 1 homeoprotein HOXA9. Nat Genet 1996; 12: 159–167. 11 Lawrence HJ, Sauvageau G, Ahmadi N, Lopez AR, Lebeau MM, Link M, Humphries K, Largman C. Stage- and lineage-specific expression of the HOXA10 homeobox gene in normal and leukemic hematopoietic cells. Exp Hematol 1995; 23: 1160–1166. 12 Celetti A, Barba P, Cillo C, Rotoli B, Boncinelli E, Magli MC. Characteristic patterns of Hox gene expression in different types of human leukemia. Int J Cancer 1993; 53: 237–244. 13 Shen W-F, Rozenfeld S, Lawrence HJ, Largman C. The Abd-B-like Hox homeodomain proteins can be subdivided by the ability to form complexes with Pbxla on a novel DNA target. J Biol Chem 1997; 272: 8198–8206. 14 Nourse J, Mellentin JD, Galili N, Wilkinson J, Stanbridge E, Smith SD, Cleary ML. Chromosomal translocation t(1;19) results in synthesis of a homeobox fusion mRNA that codes for a potential chimeric transcription factor. Cell 1990; 60: 535–545. 15 Shen W-F, Montgomery JC, Rozenfeld S, Moskow JJ, Lawrence HJ, Buchberg AM, Largman C. AbdB-like Hox proteins stabilize DNA

697


698 16

17 18 19 20

21 22

23

24

25

26

27 28

binding by the Meis1 homeodomain proteins. Mol Cell Biol 1997; 17: 6448–6458. Nakamura T, Largaespada DA, Shaughnessy JDJ, Jenkins NA, Copeland NG. Cooperative activation of Hoxa and Pbx1-related genes in murine myeloid leukaemias. Nat Genet 1996; 12: 149–153. Butler LH, Slany R, Cui X, Cleary ML, Mason DY. The HRX protooncogene product is widely expressed in human tissues and localizes to nuclear structures. Blood 1997; 89: 3361–3370. Hess JL, Yu BD, Li B, Hanson R, Korsmeyer SJ. Defects in yolk sac hematopoiesis in Mll-null embryos. Blood 1997; 90: 1799–1806. Tkachuk DC, Kohler S, Cleary ML. Involvement of a homolog of drosophila trithorax by 11q23 chromosomal translocations in acute leukemias. Cell 1992; 71: 691–700. Gu Y, Nakamura T, Alder H, Prasad R, Canaani O, Cimino G, Croce CM, Canaani E. The t(4;11) chromosome translocation of human acute leukemias fuses the ALL-1 gene, related to drosophila trithorax, to the AF-4 gene. Cell 1992; 71: 701–708. Sawyers CL, Denny CT, Witte ON. Leukemia and the disruption of normal hematopoiesis. Cell 1991; 64: 337–350. Andrews RG, Singer JW, Bernstein ID. Precursors of colony-forming cells in humans can be distinguished from colony-forming cells by expression of the CD33 and CD34 antigens and light scatter properties. J Exp Med 1989; 169: 1721–1731. Sutherland HJ, Lansdorp PM, Henkelman DH, Eaves AC, Eaves CJ. Functional characterization of individual human hematopoietic stem cells cultured at limiting dilution on supportive marrow stromal layers. Proc Natl Acad Sci USA 1990; 87: 3584–3588. Conneally E, Cashman J, Petzer A, Eaves C. Expansion in vitro of transplantable human cord blood stem cells demonstrated using a quantitative assay of their lympho-myeloid repopulating activity in nonobese diabetic-scid/scid mice. Proc Natl Acad Sci USA 1997; 94: 9836–9841. Bhatia M, Wang JCY, Kapp U, Bonnet D, Dick JE. Purification of primitive human hematopoietic cells capable of repopulating immune-deficient mice. Proc Natl Acad Sci USA 1997; 94: 5320–5325. Terstappen LWMM, Safford M, Unterhalt M, Konemann S, Zurlutter K, Piechotka K, Drescher M, Aul C, Buchner T, Hiddemann W et al. Flow cytometric characterization of acute myeloid leukemia: IV. Comparison to the differentiation pathway of normal hematopoietic progenitor cells. Leukemia 1992; 6: 993–1000. Sutherland HJ, Blair A, Zapf RW. Characterization of a hierarchy in human acute myeloid leukemia progenitor cells. Blood 1996; 87: 4754–4761. Ailles LE, Gerhard B, Hogge DE. Detection and characterization of primitive malignant and normal progenitors in patients with acute myelogenous leukemia using long-term coculture with supportive feeder layers and cytokines. Blood 1997; 90: 2555–2564.

29 Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nature Med 1997; 3: 730–736. 30 Strauss LC, Rowley SD, La Russa VF, Sharkis SJ, Stuart RK, Civin CI. Antigenic analysis of hematopoiesis. V. Characterization of My-10 antigen expression by normal lymphohematopoietic progenitor cells. Exp Hematol 1986; 14: 878–886. 31 Blair A, Hogge DE, Ailles LE, Lansdorp PM, Sutherland HJ. Lack of expression of Thy-1 (CD90) on acute myeloid leukaemia cells with long-term proliferative ability in vitro and in vivo. Blood 1997; 89: 3104–3112. 32 Brady G, Barbara M, Iscove NN. Representative in vitro cDNA amplification from individual hemapoietic cells and colonies. Meth Mol Cell Biol 1990; 2: 17–25. 33 Nakamura T, Jenkins NA, Copeland NG. Identification of a new family of Pbx-related homeobox genes. Oncogene 1996; 13: 2235–2242. 34 Eaves CJ, Cashman JD, Eaves AC. Methodology of long-term culture of human hemopoietic cells. J Tissue Cult Methods 1991; 13: 55–62. 35 Hogge DE, Lansdorp PM, Reid D, Gerhard B, Eaves CJ. Enhanced detection, maintenance and differentiation of primitive human hematopoietic cells in cultures containing murine fibroblasts engineered to produce human Steel factor, interleukin-3 and granulocyte colony-stimulating factor. Blood 1996; 88: 3765–3773. 36 Lengauer C, Green ED, Cremer T. Fluorescence in situ hybridization of YAC clones after Alu-PCR amplification. Genomics 1992; 13: 826–828. 37 Look AT. Oncogenic transcription factors in the human acute leukemias. Science 1997; 278: 1059–1064. 38 Afonja O, Smith J, Ohyashiki K, Ohyashiki J, Toyama K, Inghirami G, Takeshita K. MEIS 1 gene family in human acute myeloid leukemia. Blood 1997; 90: 67a. 39 Krosl J, Baban S, Krosl G, Largman C, Rozenfeld S, Sauvageau G. Proliferation and transformation by HOX genes involves cooperation with PBX. Blood 1997; 90: 324a. 40 Rubnitz JE, Behm FG, Downing JR. 11q23 rearrangements in acute leukemia. Leukemia 1996; 10: 74–82. 41 Lufkin T. Transcriptional control of Hox genes in the vertebrate nervous system. Curr Opin Genet Dev 1996; 6: 575–580. 42 Simon J. Locking in stable states of gene expression: transcriptional control during Drosophila development. Curr Opin Cell Biol 1995; 7: 376–385. 43 Lessard J, Baban S, Sauvageau G. Stage-specific expression of Polycomb group genes in human bone marrow cells. Blood 1998; 91: 1216–1224. 44 Kroon E, Krosl J, Thorsteinsdottir U, Baban S, Buchberg AM, Sauvageau G. Hoxa9 transforms primary bone marrow cells through specific collaboration with Meis1a but not Pbx1b. EMBO J 1998; 17: 3714–3725.