among human hematopoietic progenitor cells by ...

From bloodjournal.hematologylibrary.org by guest on June 8, 2013. For personal use only.

2004 104: 675-686 Prepublished online April 13, 2004; doi:10.1182/blood-2003-10-3423

Molecular evidence for stem cell function of the slow-dividing fraction among human hematopoietic progenitor cells by genome-wide analysis Wolfgang Wagner, Alexandra Ansorge, Ute Wirkner, Volker Eckstein, Christian Schwager, Jonathon Blake, Katrin Miesala, Jan Selig, Rainer Saffrich, Wilhelm Ansorge and Anthony D. Ho

Updated information and services can be found at: http://bloodjournal.hematologylibrary.org/content/104/3/675.full.html Articles on similar topics can be found in the following Blood collections Gene Expression (1086 articles) Genomics (149 articles) Hematopoiesis and Stem Cells (3132 articles) Information about reproducing this article in parts or in its entirety may be found online at: http://bloodjournal.hematologylibrary.org/site/misc/rights.xhtml#repub_requests Information about ordering reprints may be found online at: http://bloodjournal.hematologylibrary.org/site/misc/rights.xhtml#reprints Information about subscriptions and ASH membership may be found online at: http://bloodjournal.hematologylibrary.org/site/subscriptions/index.xhtml

Blood (print ISSN 0006-4971, online ISSN 1528-0020), is published weekly by the American Society of Hematology, 2021 L St, NW, Suite 900, Washington DC 20036. Copyright 2011 by The American Society of Hematology; all rights reserved.

From bloodjournal.hematologylibrary.org by guest on June 8, 2013. For personal use only. HEMATOPOIESIS

Molecular evidence for stem cell function of the slow-dividing fraction among human hematopoietic progenitor cells by genome-wide analysis Wolfgang Wagner, Alexandra Ansorge, Ute Wirkner, Volker Eckstein, Christian Schwager, Jonathon Blake, Katrin Miesala, Jan Selig, Rainer Saffrich, Wilhelm Ansorge, and Anthony D. Ho

The molecular mechanisms that regulate asymmetric divisions of hematopoietic progenitor cells (HPCs) are not yet understood. The slow-dividing fraction (SDF) of HPCs is associated with primitive function and self-renewal, whereas the fastdividing fraction (FDF) predominantly proceeds to differentiation. CD34ⴙ/CD38ⴚ cells of human umbilical cord blood were separated into the SDF and FDF. Genomewide gene expression analysis of these populations was determined using the newly developed Human Transcriptome

Microarray containing 51 145 cDNA clones of the Unigene Set-RZPD3. In addition, gene expression profiles of CD34ⴙ/CD38ⴚ cells were compared with those of CD34ⴙ/ CD38ⴙ cells. Among the genes showing the highest expression levels in the SDF were the following: CD133, ERG, cyclin G2, MDR1, osteopontin, CLQR1, IFI16, JAK3, FZD6, and HOXA9, a pattern compatible with their primitive function and self-renewal capacity. Furthermore, morphologic differences between the SDF and FDF were determined. Cells in the

SDF have more membrane protrusions and CD133 is located on these lamellipodia. The majority of cells in the SDF are rhodamine- 123dull. These results provide molecular evidence that the SDF is associated with primitive function and serves as basis for a detailed understanding of asymmetric division of stem cells. (Blood. 2004;104:675-686)

© 2004 by The American Society of Hematology

Introduction Stem cells are characterized by their dual abilities to self-renew and to differentiate into progenitors of various lineages. Using hematopoietic stem cells (HSCs) as a model, these characteristics have been used to define stem cells for over 40 years, but the knowledge of the processes involved in regulating self-renewal versus differentiation is still only rudimentary.1 Several surrogate markers and assay systems have been developed, but all in vitro systems have failed to definitively identify the ultimate HSCs. The CD34⫹/ CD38⫺ immunophenotype seems to be associated with a primitive population in human bone marrow and cord blood.2,3 Further enrichment of HSCs might be achieved by selection for other phenotypic markers such as Thy-1, HLA-DR, CD133, c-kit, or rhodamine-123dull.4-8 However, there is no appropriate phenotypic constellation that permits us to isolate a pure and homogenous population of HSCs. Provided with knowledge gained in genomics in the past few years, gene expression analysis of hematopoietic progenitor cells (HPCs) could help to reveal the molecular mechanisms that are involved in self-renewal or differentiation.9-15 To determine the distinct molecular characteristics of multipotent HPCs, it is necessary to enrich pure cell populations with the dual ability of self-renewal and differentiation.15 This dual ability implies that stem cells undergo at some stage during the development asymmetric divisions, resulting in daughter cells with different long-term fate. Previous work of our and other groups has shown that

asymmetric cell division of HPCs coincided with primitive function. Using time-lapse microscopy recording and continuous image analysis systems, we have correlated the symmetry of HPC divisions with their long-term fate at a single-cell level. We observed that about 30% of CD34⫹/CD38⫺ cells, irrespective of source (ie, from fetal liver, umbilical cord blood, or adult bone marrow), gave rise to a daughter cell that remained quiescent over 8 days, whereas the other daughter cell proliferated exponentially.16 At a single-cell level, the primitive myeloid-lymphoid–initiating cells (ML-ICs) were only found in quiescent or slow-dividing fractions (SDFs).17,18 Other authors have also demonstrated that most primitive stem cells were quiescent or had low cycling rates,19,20 and that most of the CD34⫹/CD38⫺ cells analyzed directly after their isolation from the bone marrow remained quiescent.21 In the nonobese diabetic/severe combined immunodeficient (NOD/SCID) mouse model, quiescent cells residing in G0 had a significantly higher repopulating capacity than the proliferating fraction.22,23 Thus, separation of the slow-dividing subfraction within the CD34⫹/CD38⫺ population could further enrich cells associated with primitive function and asymmetric division. We and others have demonstrated that the membrane dye PKH26 could be exploited to assist in isolation of the quiescent or SDF.16,24-26 In continuation with this line of research, the CD34⫹/CD38⫺ cells were separated into an SDF and a fast-dividing fraction (FDF). Global gene expression profiles for these 2 populations

From the Department of Medicine V, University of Heidelberg, Heidelberg, Germany; and Biochemical Instrumentation Programme, European Molecular Biology Laboratory, Heidelberg, Germany.

An Inside Blood analysis of this article appears in the front of this issue.

Submitted October 7, 2003; accepted March 24, 2004. Prepublished online as Blood First Edition Paper, April 13, 2004; DOI 10.1182/blood-2003-10-3423. Supported by Deutsche Forschungsgemeinschaft (DFG) HO 914/2-1, HO 914/3-1, Bundesministerium fu¨r Bildung und Forschung (BMBF) 01GN0107, and Siebeneicher Stiftung, Germany.

BLOOD, 1 AUGUST 2004 䡠 VOLUME 104, NUMBER 3

Reprints: Anthony D. Ho, Department of Medicine V, University of Heidelberg, Hospitalstrasse 3, 69115 Heidelberg, Germany; e-mail: [email protected]. The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked ‘‘advertisement’’ in accordance with 18 U.S.C. section 1734. © 2004 by The American Society of Hematology

675

From bloodjournal.hematologylibrary.org by guest on June 8, 2013. For personal use only. 676


WAGNER et al

were analyzed and compared using the recently developed novel Human Transcriptome Microarray (U.W., Christian Maercher, Mechthild Wagner, J. S., Heiko Drzoriek, Uwe Radelof, A. A., C. S., Bernhard Korn, and W. A., manuscript in preparation) with 51 145 UnigeneSet-RZPD3 cDNA clones. To obtain additional information on primitive versus committed progenitor cells, the gene expression profile of CD34⫹/CD38⫺ cells, which are associated with more primitive function, was analyzed and compared with that of the CD34⫹/CD38⫹ cells that are committed to differentiation.

Material and methods Cell isolation Human cord blood (CB) was collected from the umbilical cord after informed consent was obtained from the patient, using guidelines approved by the Ethics Committee on the Use of Human Subjects at the University of Heidelberg. Mononuclear cells (MNCs) were isolated after centrifugation on Ficoll-Hypaque (Biochrom, Berlin, Germany). CD34⫹ cells were enriched with a monoclonal anti-CD34 antibody labeled with magnetic beads on an affinity column (Miltenyi Biotec, Bergisch-Gladbach, Germany). CD34⫹-enriched cells were incubated with anti–CD34-phycoerythrin (PE; Becton Dickinson, San Jose, CA) and anti–CD38-allophycocyanin (APC; Becton Dickinson). The cells were washed in phosphate-buffered saline (PBS) and 10% fetal calf serum (FCS) and stained with propidium iodide (PI) to identify dead cells. CD34⫹/CD38⫺ and CD34⫹/CD38⫹ populations were sorted using the automatic cell-depositing unit on a fluorescence-activated cell sorting (FACS)–Vantage-SE flow cytometry system (Becton Dickinson, San Jose, CA). Cells positive for PI were excluded. The subpopulations were either used directly for gene expression profiling or labeled with the membrane dye PKH26 (Sigma, St Louis, MO) to monitor cell proliferation. The labeled cells were cultured for 7 days in Myelocult (Stem Cell Technology, Vancouver, BC, Canada) in 24-well plates (Nunc, Roskilde, Denmark). The medium was supplemented with 2.5 U/mL recombinant human erythropoietin (Roche, Hertfordshire, United Kingdom), 10 ng/mL recombinant human interleukin 3 (IL-3), 500 U/mL recombinant human IL-6, 10 ng/mL recombinant human granulocyte-macrophage colony-stimulating factor, 2.5 ng/mL recombinant human basic fibroblast growth factor, 10 ng/mL recombinant human insulin-like growth factor 1, 50 ng/mL recombinant human stem cell factor, 2.5 ng/mL fibroblast growth factor ␤, 1000 U/mL penicillin, and 100 U/mL streptomycin (Gibco, Grand Island, NY) as described previously.16 Cytokines were purchased from R&D Systems (Minneapolis, MN). Cells were resorted according to their PKH26 staining after 7 days. RNA isolation and probe synthesis About 5 ⫻ 104 cells from each fraction were lysed and total RNA isolated using the Acturus PicoPure RNA isolation kit (Acturus, Mountain View, CA). DNase treatment was performed (Quiagen, Hilden, Germany). RNA quality was controlled with the RNA 6000 Pico LabChip kit (Agilent, Waldbronn, Germany). Linear amplification of mRNA was performed by in vitro transcription over 2 rounds using the Arcturus RiboAmp kit (Acturus). The aRNA was analyzed by the RNA nano LabChip kit (Agilent) and by the SpectraMAX plus photometer (Molecular Devices, Sunnyvale, CA) at 260 nm. We obtained about 130 ␮g aRNA with a continuous spectrum of RNA length of 200 to 700 bp (results not shown). About 25 ␮g aRNA was incubated with 3 ␮g random primer (Invitrogen, Karlsruhe, Germany) and labeled by amino-allyl-coupling using the Atlas Glass Fluorescent Labeling Kit (Clontech, Palo Alto, CA) and Cy3/Cy5-monofunctional reactive dye (Amersham Biosciences, Little Chalfont, United Kingdom).

The Human Transcriptome Microarray A novel cDNA microarray was used that represents the UnigeneSet-RZPD3 composed of 51 145 clones, a very well-characterized subset of the IMAGE cDNA clone collection (http://image.llnl.gov/image). Sequence information and clones are available at the Resource Center and Primary Database (RZPD, Berlin, Germany). Selection of the clones was based on the human Unigene clusters of the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db ⫽ unigene). The inserts are selected to be located close to the 3⬘ end of the cDNA and to have a length between 500 and 1200 bp. All clones have been sequence verified. A total of 42 000 of the 45 000 successfully resequenced clones are associated with ENSEMBL or LocusLink gene predictions. The cDNA inserts were amplified with a pair of NH2-modified flanking universal primers. The polymerase chain reaction (PCR) products were purified on a Biomek FX Robot (Beckman Coulter, Krefeld, Germany) using the Macherey-Nagel Nucleospin kit. The purified fragments were analyzed on a buffer-free ready-to-run agarose gel system (RTR; Amersham/Pharmacia, Freiburg, Germany). Gels were loaded by a Biomek 2000 robot (Beckman Coulter) and band patterns analyzed by the EMBL “ChipSkipper” software (C. Schwager, EMBL, Heidelberg, Germany). Over 96% of the inserts could be amplified and showed a single band pattern. The spotting was performed on a GeneMachines OmniGrid spotter with 48 split needles (Arraylt, Sunnyvale, CA) depositing the clone set on 2 epoxy slides. Further details about the techniques used have been described before.27,28 Hybridization of the slides The slides were immersed at 42°C in 6 ⫻ standard saline citrate (SSC)/ 0.5% sodium dodecyl sulfate (SDS)/1% bovine serum albumin (BSA) for 40 minutes and carefully washed in ddH2O at room temperature. The attached PCR products were then denatured at 95°C in ddH2O for 2 minutes and then air dried. Prior to hybridization, the purified Cy3/Cy5-labeled cDNA probes were mixed together. Then, 20 ␮g poly-d-A and 20 ␮g human Cot1DNA (both Gibco Invitrogen, Carlsbad, CA) were added and the mixture evaporated in the Vacuum Concentrator 5301 (Eppendorf, Hamburg, Germany) at 60°C to complete dryness. The pellet was dissolved in 50 ␮L hybridization buffer (50% formamide/6 ⫻ SSC/0.5% SDS/5 ⫻ Denhardt) and denatured by incubating at 95°C for 2 minutes. The probe was hybridized with the microarray in a humid chamber in a water bath at 42°C for 16 hours. After hybridization, slides were washed in 0.1 ⫻ SSC/0.1% SDS for 10 minutes and then twice with 0.1 ⫻ SSC for 5 minutes at 37°C with 130 rpm on an orbital shaker (Gio Gyrotory Shaker, New Brunswick Scientific, Edison, NJ). Slides were dried by a brief centrifugation at 715 g in a microtiter plate centrifuge (Z320; Hermle, Wehingen, Germany) and scanned immediately. Analysis of results Slides were scanned at a 10-␮m resolution using the GenePix 4000B Microarry-Scanner (Axon Instruments, Union City, CA). Individual laser power and photomultiplier settings were used, allowing all signals to remain in the linear range of the scanner. Separate images for Cy3 and Cy5 were produced and analyzed by the ChipSkipper Microarray Data Evaluation Software (http://chipskipper.embl.de). Integration grids were automatically centered to the images. For each spot total intensity and local background were calculated. Raw ratios were derived from background reduced signals. Normalization was performed by intensity-dependent windowed median ratio centering. The resulting data were further analyzed by Excel (Microsoft, Redmond, WA). The ratio of differential gene expression is presented as the mean of Log2 values from all replicate and color-flip hybridizations (Log2 ratio). Genes with mean Log2 expression greater than 1 were selected as those that show a significant expression ratio across the replicated data sets. Further filtering was performed with replica standard deviation (SD). We have considered several methods for the statistical analysis including statistical analysis for microarrays (SAMs), which was developed to analyze a treated versus an untreated set (as a reference) with single-color arrays.29 For our experimental design we have

From bloodjournal.hematologylibrary.org by guest on June 8, 2013. For personal use only. BLOOD, 1 AUGUST 2004 䡠 VOLUME 104, NUMBER 3

chosen the 2-tailed, paired Student t test to estimate the probability of each filtered gene to be part of the broad distribution of all genes in the 4 (CD34⫹CD38⫺/CD34⫹CD38⫹) or 6 data sets (SDF/FDF). False discovery rate (FDR) was estimated by simulations. Using a pure gaussian distributed set of replicated hybridizations (generated with Monte Carlo methods) no false-positive genes could be detected with our filtering method. Taking into account that real data sets contain a biologic bias, a second simulation was performed. Stochastic permutations of all experimental ratio values from the replicated hybridizations were used to create sets of experimental but not correlated virtual replications. Ten thousand simulations were performed and the average number of genes within the filter criteria was given as FDR. The complete microarray data including the description of all spotted genes (according to Minimal Information About Microarray Experiments, MIAME requirements30) is submitted to the public microarray database ArrayExpress (http://www.ebi.ac.uk/arrayexpress/query/entry, accession number E-EMBL-1). Comparison with other databases Murine and human sequences were compared in a similar way as described previously.9 Homologous genes were identified by identical Unigene names. Then sequences from the UnigeneSet-RZPD3 clone set were blasted against release 8 of the mouse Database of Transcription (http:// www.cbil.upenn.edu/downloads/DoTS) using an e-value cutoff of 1e-15. Identified sequences were further verified by tblastx using an e-value cutoff of 1e-10. Accession numbers from DoTS transcript were used to map to Unigene clusters. These clusters were cross-referenced against mouse genes, reported to be enriched in the stem cell fractions.

FUNCTION OF SLOW-DIVIDING FRACTION IN HPCs

677

Rhodamine assay Efflux of the fluorescent dye rhodamine-123 (Rh-123) resolves functionally distinct subsets of HSCs.32,33 Cells were washed and resuspended in growth medium as described (see “Cell isolation”) containing 0.2 ␮g/mL Rh-123 (Sigma-Aldrich, Deisenhofen, Germany). After an incubation period of 30 minutes, cells were centrifuged and resuspended in growth medium (see “Cell isolation”). Cells were incubated at 37°C for 60 minutes to allow the efflux of Rh-123. Then the cells were washed twice in PBS and 10% FCS and analyzed by flow cytometry. Additional controls were performed with 1.5 ␮M of the P-glycoprotein inhibitor cyclosporin A (Novartis, Basel, Switzerland) in all steps. Immunofluorescence An Olympus IX70 microscope (Olympus Optical, Hamburg, Germany) was used to assess the fluorescence and morphology of the PKH26⫹/⫺ cells. For simultaneous analysis of CD133 and PKH26, the antibodies CD133/1 pure antibody (Miltenyi Biotec) and the Alexa Fluor-488 antimouse secondary antibody (Molecular Probes, Eugene, OR) were used, and analysis was performed with a dual-band filter fluorescein isothiocyanate (FITC)/Cy3 (AHF Analysetechnik, Tu¨ bingen, Germany). Artificial capping of the antigen to membrane protrusions was excluded by additional fixation with 4% formaldehyde. The proportion of irregular cells with membrane protrusions was calculated by random choice of 5 different fields (⫻ 20 objective) of each condition and direct counting of total cells (400-500).

Results Isolation of HPCs

RT-PCR analysis To further support the information obtained from the microarray data, real-time reverse transcription-PCR (RT-PCR) analysis was carried out with LightCycler technology (Roche Molecular Biochemicals, Mannheim, Germany). About one twentieth of total RNA isolated as described (see “RNA isolation and probe synthesis”) was reverse transcribed by Superscript II (Gibco Invitrogen). Semiquantitative PCR was performed with the LightCycler Master SYBR Green kit (Roche Molecular Biochemicals) with 3 mM MgCl at 480 seconds of preincubation at 95°C followed by 50 cycles of 10 seconds at 55°C, 25 seconds at 72°C, and 10 seconds at 95°C. PCR products were subjected to melting curve analysis and to conventional agarose gel electrophoresis to exclude synthesis of unspecific products. The following primers were used: 18s rRNA primers supplied by Ambion (Austin, TX): 18s rRNA forward primer: 5⬘-TCAAGAACGAA AGTCGGAGG-3⬘; 18s rRNA reverse primer: 5⬘-GGACATCTAA GGGCATCACA-3⬘; GAPDH forward primer: 5⬘-ATGGCACCGT CAAGGCTGAG A-3⬘; GAPDH reverse primer: 5⬘-GGCATGGACT GTGGTCATGA G-3⬘; ubiquitin forward primer: 5⬘-TGCAGAGTAA TGCCATCACTG-3⬘; ubiquitin reverse primer: 5⬘-CAAGATAAAG AAGGCATCCC C-3⬘; CD133 forward primer: 5⬘ACATGAAAAG ACCTGGGGG-3⬘; CD133 reverse primer: 5⬘-GATCTGGTGT CCCAGCATG-3⬘; HLA-A forward primer: 5⬘-ACGACGGCAA GGATTACATC-3⬘; HLA-A reverse primer: 5⬘-GCTTCATGGT CAAGAGACAG C-3⬘; ␥-globin forward primer: 5⬘-TTGAAAGCTC TGAATCATGG G-3⬘; ␥-globin reverse primer: 5⬘-CTTCAAGCTC CTGGGAAATG-3⬘; CD36 forward primer: 5⬘-TTTATGAGGC GATTATGACAG C-3⬘; CD36 reverse primer: 5⬘-AGTTGCAACT TACGCTTGGC-3⬘; cadherin 1 forward primer: 5⬘-TGTTTTCCTT TTCCACCCC-3⬘; and cadherin 1 reverse primer: 5⬘-ACCCTGCAAT CACTTTTTGG-3⬘. Primers were designed within the sequences represented on the Human Transcriptome Microarray and synthesized by Biospring (Frankfurt, Germany). The amplification efficiency of PCR products was determined by calculating the slope after semilogarithmic plotting of the values against cycle number as described before.31 At least 3 independent total RNA samples were analyzed and differential expression calculated in relation to 18s rRNA.

The CD34⫹/CD38⫺ fraction and the CD34⫹/CD38⫹ fraction were isolated from human umbilical CB. We then separated the CD34⫹/ CD38⫺ population into an SDF and an FDF. CD34⫹/CD38⫺ cells were labeled with the fluorescent dye PKH26 to assist this separation. During 1 week of in vitro cultivation in the medium, this dye was diluted after every cell division and distributed equally to the daughter cells. The SDF could then be isolated as PKH26⫹ and the FDF as PKH26⫺. Reanalysis of the isolated cells revealed a purity of 95% to 99% in both populations (Figure 1). Previous experiments from our group have demonstrated that approximately 30% of the CD34⫹/CD38⫺ cells divide slowly and asymmetrically and this divisional behavior is associated with primitive function.16,18 Gene expression profiles: CD34ⴙ/CD38ⴚ versus CD34ⴙ/CD38ⴙ

The CD34⫹/CD38⫺ fraction is enriched in primitive HPCs. In our experiments the gene expression profile of this population was determined in comparison to the more committed CD34⫹/CD38⫹ population. For each population, 2 independent RNA pools were analyzed. Each of these RNA pools was a combination of 3 total RNA samples isolated from 3 individual CBs pooled prior to amplification. Direct and corresponding color-flip experiments were carried out to compensate for any dye-specific effects, resulting in a total of 4 cohybridization data sets. Hybridizations of the Human Transcriptome Microarray with RNA samples from CD34⫹/CD38⫺ and CD34⫹/CD38⫹ cells resulted in reproducible signal intensities and expression patterns. Ninety-six spots showed a more than 2-fold (Log2 ratio ⬎ 1, SD ⬍ 1, FDR ⫽ 12) higher signal in the CD34⫹/CD38⫺ stem cell fraction. Among these were 27 functionally characterized genes summarized in Table 1. In contrast, 119 spots showed at least a 2-fold (Log2 ratio ⬍ ⫺1, SD ⬍ 1, FDR ⫽ 27) higher signal in the CD34⫹/CD38⫹ population. Among these were 38 functionally characterized genes (Table 2).



WAGNER et al

Figure 1. Separation of CD34ⴙ/CD38ⴚ cells into SDF and FDF. CD34⫹/CD38⫺ and CD34⫹/CD38⫹ subpopulations were isolated from human CB (A). Reanalysis of the sorted cells confirmed the efficient separation of these cell fractions (B-C). The CD34⫹/ CD38⫺ stem cell population was then stained with PKH26 and cultivated for 1 week. The SDF (PKH26⫹) was separated by cell sorting form the FDF (PKH26⫺) population (D). Efficient enrichment was controlled by reanalysis (E-F).

Gene expression profile: SDF versus FDF

The SDF within the CD34⫹/CD38⫺ population is associated with asymmetric cell division and multipotent ML-IC activity. We have determined the gene expression profile of the SDF in comparison to the FDF for 3 individual CB samples separately. Corresponding color-flip experiments were performed, resulting in 6 cohybridization data sets. Global gene expression survey by scatter plot analysis revealed that a large proportion of genes was differentially regulated in the 2 populations. Comparative analysis with respect

to the CD34⫹/CD38⫺ versus CD34⫹/CD38⫹ populations showed repeatedly additional heterogeneity of signals. This corresponds most probably to the greater biologic differences between the SDF and FDF populations (Figure 2). In the SDF, a total of 942 spots showed at least a 2-fold higher signal (Log2 ratio ⬎ 1, FDR ⫽ 39). The 36 genes that were more than 4-fold overexpressed in the SDF are summarized in Table 3 (Log2 ratio ⬎ 2, FDR ⫽ 0). Many genes of the HLA cluster, the homeodomain protein HOXA9 (3.9-fold),

Table 1. Genes up-regulated more than 2-fold in the CD34ⴙ/CD38ⴚ fraction versus CD34ⴙ/CD38ⴙ Gene

Symbol

Accession no.

Ratio (Log2)

SD

Cell surface receptors/ligands Similar to endothelial cell-selective adhesion molecule*

ESAM

R64386

1.1

0.3

s-lac lectin 1-14-ii (lgals2)†

LGALS2

AA872397

1.2

0.8

Endomucin-2*

LOC51705

AA426155

1.3

0.3

Insulin-like growth factor 2 (igf-2)†

IGF2

H59614

1.0

0.6

AI819700

1.2

0.3

N52513

1.1

0.2

W65323

1.1

0.8 0.4

Transporter/signal transduction ras-related protein r-ras† sh3 binding protein*

SH3BP5

ESTs, similar to ICE4 human caspase-4 Osteopontin†

SPP1

R59298

1.0

mRNA for silencer element†

STMN2

R19072

1.0

0.5

Potassium channel, subfamily k, member 17 (task-4)†

KCNK17

AI740592

1.3

0.4

fxyd domain-containing ion transport regulator 6*

FXYD6

R87333

1.2

0.2

PR-domain zinc finger protein 1†

PRDM1

AI692550

1.0

0.3

Transcription factors/cell cycle Sperm-associated antigen 9

SPAG9

AI690418

1.2

0.8

Homo sapiens mRNA for rbp-ms/type 3†

RBPMS

AA045722

1.3

0.4

Histones h2b.1 and h2a†

H2BFQ

R98471

1.1

0.4

Cytokeratin 20†

KRT20

Aa471027

1.3

0.7

Human profilin ii†

PFN2

R17470

1.1

0.5

Ankyrin g (ank-3)*

ANK3

H19467

1.5

0.2

Pancreatic trypsinogen (try2)*

PRSS2

AA293076

1.8

0.3

Ad036 protein‡

LOC51313

R15218

1.8

0.2

Hypothetical protein flj10697*

FLJ10697

H80748

1.0

0.2

Selenoprotein p†

—

N78730

1.0

0.5

kiaa0587 protein*

NCKAP1

R32815

1.1

0.5

Hypothetical protein smap31†

SMAP31

AA504137

1.2

0.3

kiaa1497 protein*

KIAA1497

R55372

1.4

0.3

kiaa1573 protein†

KIAA1573

H48270

1.2

0.6

Cellular component/cytoskeleton

Other/unknown

The Log2 ratios of (CD34⫹/CD38⫺) to (CD34⫹/CD38⫹) were the average values of 4 hybridizations in 2 experiments. Error estimate is given as SD. Probability of differential gene expression in relation to all high-quality signals was determined with the Student t test. Accession numbers from GenBank are provided. *P ⬍ .005. †P ⬍ .05. ‡P ⬍ .0005.



679

Table 2. Genes down-regulated more 2-fold in the CD34ⴙ/CD38ⴚ fraction versus CD34ⴙ/CD38ⴙ Gene

Symbol

Accession no.

Ratio (Log2)

SD

Cell surface receptors/ligands CD38 antigen*

CD38

AA417096

⫺1.9

0.5

Kell blood group protein

KEL

H61054

⫺1.1

0.8

IL-7 receptor

IL7

N95337

⫺1.5

0.9

CD24 antigen†

CD24

R96738

⫺1.8

1.6

Antigen of monoclonal antibody ki-67†

MKI67

T72662

⫺1.3

0.6

Colony-stimulating factor 2 receptor ␤*

CSF2RB

H04648

⫺1.8

0.3

T-cell receptor ti rearranged ␥-chain mRNA V-J-C region†

TRG

R77295

⫺2.1

0.9

CD36 antigen†

CD36

R09416

⫺1.2

0.4

Ig rearranged ␥ chain mRNA, V-J-C region†

IGKC

R10121

⫺1.2

0.3

High-affinity IgE receptor ␣ subunit (fceri)†

FCER1A

AI676097

⫺1.1

0.5

Ig-related 14.1 protein†

IGLL1

W73587

⫺1.2

0.5

Hepatocyte growth factor (HGF)†

HGF

R37604

⫺1.0

0.6

Macrophage inflammatory protein-1␣/RANTES receptor†

CCR1

AI570752

⫺1.1

0.4

Solute carrier family 11†

SLC11A3

T97261

⫺1.2

0.5

5-Lipoxygenase activating protein (FLAP)†

ALOX5AP

H02307

⫺2.1

0.8

Ionotropic ATP receptor p2⫻5a mRNA‡

P2RX5

AF070573

⫺1.2

0.0

Very-long-chain acyl-CoA synthetase*

FACVL1

AA417046

⫺1.2

0.2

L-histidine decarboxylase*

HDC

AI695885

⫺3.7

0.7

Lymphoid enhancer-binding factor-1†

LEF1

W81128

⫺1.6

0.7

Cyclin PCNA*

PCNA

T82974

⫺1.0

0.1

Cyclin B2*

CCNB2

AA129846

⫺1.3

0.3

Erythroid Kruppel-like factor (EKLF)*

KLF/1

H65733

⫺1.6

0.3

AA437223

⫺1.2

0.7

GDF8

H92027

⫺1.1

0.6

Pleckstrin (p47)†

PLEK

R05932

⫺1.0

0.4

Mast cell carboxypeptidase a (MC-CPA)*

CPA3

N79582

⫺1.9

0.4

Light and heavy chains of myeloperoxidase*

MPO

R05801

⫺3.3

0.5

Kinesin-like 7*

KNSL7

AI623136

⫺1.5

0.4

Kinesin-like 5 (mitotic kinesin-like protein 1)*

KNSL5

AI288408

⫺1.0

0.2

kiaa0473 protein*

DNAJC6

N38951

⫺1.1

0.2

hspc037 protein*

LOC51659

H58066

⫺1.4

0.3

kiaa0101 gene*

KIAA0101

R36629

⫺1.6

0.4

Retinol dehydrogenase homolog†

RDHL

AI540484

⫺1.7

0.5

N-acetylglucosamine-6-o-sulfotransferase (glcnac6st)†

CHST2

N80041

⫺1.1

0.3

Protein associated with glycosphingolipid microdomains†

PAG

N50114

⫺1.3

0.4

Hypothetical protein flj10540*

FLJ10540

AI696847

⫺1.2

0.3

zw10 interactor zwint†

ZWINT

AA292765

⫺1.2

0.4

Factor XIII subunit a†

FI3A1

H67759

⫺1.4

0.6

Hyaluronan receptor (rhamm)†

HMMR

AA460915

⫺1.5

0.8

Transporter/signal transduction

Transcription factors/cell cycle

ESTs, weakly similar to nucleolin† Homo sapiens myostatin (gdf8)† Cellular component/cytoskeleton

Other/unknown

The Log2 ratios of (CD34⫹/CD38⫺) to (CD34⫹/CD38⫹) were the average values of 4 hybridizations in 2 experiments. Error estimate is given as SD. Probability of differential gene expression in relation to all high-quality signals was determined with the Student t test. ATP indicates adenosine triphosphate; acyl-CoA, acyl-coenzyme A. *P ⬍ .005. †P ⬍ .05. ‡P ⬍ .0005.

frizzled 6 (2.2-fold), and JAK3 (2.3-fold) were also highly expressed in the SDF. A total of 794 spots showed a more than 2-fold higher signal in the FDF (Log2 ratio ⬍ ⫺1, FDR ⫽ 61). Table 4 displays the 18 genes that had a more than 4-fold higher expression in the FDF (Log2 ratio ⬍ ⫺2, FDR ⫽ 0). Combination of gene expression profiles

The differential gene expression profiles of CD34⫹/CD38⫺ versus CD34⫹/CD38⫹ cells and of SDF versus FDF were compared. A total of 55 spots showed at least a 1.8-fold higher expression in both more primitive progenitor cell fractions (CD34⫹/CD38⫺ and

SDF, Log2 ratio ⬎ 0.85, FDR ⫽ 3), whereas 45 spots showed higher expression in the 2 more committed fractions (CD34⫹/ CD38⫹ and FDF, Log2 ratio ⬍ ⫺0.85, FDR ⫽ 2). Tables 5 and 6 summarize a selection of those genes. Various groups have used microarray approaches before to determine gene expression profiles in different fractions of murine and human HSCs. To further narrow down the candidate genes that are predominantly expressed in HSCs, we compared our results to 3 microarray approaches9,10,15 and analyzed the overlap (Table 7). A global comparison of the results is difficult due to differences in starting cell material, in methods (cell preparations, RNA amplification, RNA labeling, microarray design, and hybridization



WAGNER et al

the SDF especially on the tip of plastic adherent lamellipodia. In contrast, the FDF showed less podia formation and low prominin expression (Figure 4). The fluorescent dye Rh-123 can resolve functionally distinct subsets of HSCs. The majority of cells in the SDF were able to efflux Rh-123 in contrast to most cells in the FDF that remained stained with Rh-123. Under exposure to the P-glycoprotein inhibitor cyclosporin (CsA), this efflux effect was blocked (Figure 5).

Discussion

Figure 2. Global gene expression survey with the Human Transcriptome Microarray. The Human Transcriptome Microarray is composed of over 51 145 ESTs of the UnigeneSet-RZPD3. This set of PCR-amplified clones was spotted on 2 slides. The scatter plots show the relative signal intensities of the red and green channel of representative experiments. This allows global survey of differential gene expression. Lines indicate 2-, 4-, and 8-fold difference in the relative signal intensity. The 2 corresponding scatter plots represent the 2 parts of the Human Transcriptome Microarray of the same hybridization experiment. The observed differential gene expression varied more in the SDF versus FDF in comparison to CD34⫹/CD38⫺ versus CD34⫹/CD38⫹.

methods), and in the annotation of genes and data analysis. However, several genes that were up-regulated in the CD34⫹/ CD38⫺ fraction or in the SDF have also been up-regulated in other fractions of HSCs. Among the genes that were prominently overexpressed in these studies were HOXA9, frizzled 6 (FZD6), MDR1, RNA-binding protein with multiple splicing (RBPMS), PR-domain zinc finger protein 1 (PRDM1), and Janus kinase 3 (JAK3). Data supporting the microarray results

We have validated the expression of several genes in the FDF and SDF by semiquantitative RT-PCR. In accordance to the microarray analysis, GAPDH and ubiquitin showed no differential expression in these populations; CD133 and HLA-A showed higher expression in the SDF, whereas ␥-globulin, CD36, and e-cadherin (RT-PCR product only in FDF) were more highly expressed in the FDF. As expected, CD38 shows a higher expression in the CD34⫹/CD38⫹ population in microarray analysis (4.0 ⫾ 1.2-fold) and in flow cytometry (14.3 ⫾ 1.1-fold). We further analyzed the expression of CD34 and CD38 in the SDF and FDF by flow cytometry. In accordance with our microarray results the SDF revealed higher expression of CD34 (FACS, 3.4-fold; microarray, 1.4 ⫾ 0.5-fold) and of CD38 (FACS, 3.8-fold; microarray, 3.0 ⫾ 1.7-fold). Both the RT-PCR as well as the FACS techniques confirmed our microarray results on the genes studied (Figure 3). Slow-dividing stem cells have distinct morphology

Microscopic observation and flow cytometry demonstrated that cells of the SDF appeared to be smaller than those in the FDF. In the SDF the majority of cells (69% ⫾ 3%) displayed a fibroblastoid, elongated phenotype, whereas in the FDF the majority of cells have a round phenotype and only 16% ⫾ 7% showed membrane protrusions (observed in ⬎ 9 independent experiments). AntiCD133 monoclonal antibody revealed expression of prominin in

In this study we have described the transcriptional and cellular characteristics of a hematopoietic progenitor population that has been enriched by exploiting differential division kinetics. Genomewide analysis supports the notion that the SDF of CD34⫹/CD38⫺ cells is associated with primitive stem cell function. To fulfill their dual function of self-renewal as well as differentiation into progenitors of multiple blood cell lineages, HPCs must undergo asymmetric divisions during development to sustain long-term hematopoiesis as well as to produce the various progeny cells of the different lineages. In previous experiments we have monitored early cell divisions of HPCs and related them directly to long-term fate of the daughter cells at a single-cell level, confirming that asymmetric division kinetics correlates with primitive stem cell function.16,18,34 Cells giving rise to primitive ML-ICs demonstrated significantly slower division kinetics than those giving rise to committed progenitors. Furthermore, it was shown that maintenance of self-renewal and primitive function could only be influenced by direct contact with AFT024-feeder layer, which elevated the proportion of cells undergoing asymmetric divisions.34 To identify the genes involved in asymmetric division a Human Transcriptome Microarray was used. This microarray allowed the expression analysis of nearly all human transcripts isolated thus far. The complete human genome contains an estimated 30 000 to 40 000 different genes. Only about 20 000 genes are allocated and an even smaller number is functionally characterized today.35 The Human Transcriptome Microarray used is to our knowledge the largest cDNA microarray today (51 145 expression sequence tags [ESTs]) that is estimated to represent about 95% of human genes with a well-characterized clone selection.36,37 The results of gene expression analysis were confirmed by corresponding semiquantitative RT-PCR results for several selected genes as well as by immunofluorescence and flow cytometry. This new powerful tool will facilitate the identification of relevant and differentially expressed genes and gene families that are involved in fundamental life processes. In a first step, the gene expression profiles of relatively primitive CD34⫹/CD38⫺ cells versus committed CD34⫹/CD38⫹ cells were compared. As derived from the similarities in gene expression profiles, the CD34⫹/CD38⫺ and CD34⫹/CD38⫹ fractions represent closely related samples. Several genes that may be implicated in the functional organization of primitive HPCs were overexpressed in the CD34⫹/CD38⫺ fraction: insulin-like growth factor 2 (IGF2), MDR1, MN1, STMN2, ESAM 1, endomucin-2, and galectin-1. IGF2 is associated with proliferation and differentiation while maintaining a greater number of progenitor cells.38 MDR1 is a multiple drug efflux pump that is responsible for elimination of Rh-123 and has been reported to be characteristic for HSCs.39 MN1 is a tumor suppressor gene involved in an acute form of myeloid leukemia through translocation.40 The silencer element STMN2 is



681

Table 3. Genes up-regulated more than 4-fold in SDF versus FDF Gene

Symbol

Accession no.

Ratio (Log2)

SD

Cell surface receptors/ligands HLA-DR ␣ heavy-chain a class II antigen (MHC)*

HLA-DRA

H78841

2.5

0.8

Peripheral lymph node homing receptor homolog*

SELL

H00662

3.3

1.3

MHC class II dr ␤*

HLA-DRB1,5

H90289

2.0

0.9

CD133 antigen†

PROML1

R40057

3.7

0.8

m130 antigen cytoplasmic variant 2*

CD163

N73867

3.2

1.7

HLA-DP-␤ 1 gene and HLA-DP-␣-1 gene exon 1‡

HLA-DPB1

AA496472

2.2

0.7

Complement component 1, q subcomponent, receptor 1*

CIQR1

H28298

2.1

1.6

Human platelet endothelial cell adhesion molecule‡

PECAM1

R22412

2.2

0.6

Low-affinity IgG receptor CD16 (fcgriii)*

FCGR3A

N47495

2.7

1.4

Toll/IL-1 receptor-like protein 4 (til4)

TLR2

AA004805

2.1

1.6

Mannose receptor, c type 1*

MRC1

AA043388

2.2

1.0

HLA-SB(dp) ␣ gene*

HLA-DPA1

H38801

2.0

0.5

Multidrug resistance gene 1 (MDR1)‡

ABCB1

AA135957

2.1

0.5

Soluble guanylate cyclase small subunit*

GUCY1B3

W73155

2.0

1.4

Glucose transport-like 5 (glut5)*

SLC2A5

AA459707

2.0

1.4

Osteopontin*

SPP1

R59298

3.0

1.8

Interferon ␥–induced protein*

IFI16

AA126928

2.0

1.0

Prepromultimerin

MMRN

AA423867

2.1

1.4

kiaa0337 gene*

KIAA0337

AA236696

2.0

0.7

v-ets erythroblastosis virus e26 oncogene-like (avian)‡

ERG

R01192

2.9

0.9

erg2 gene†

ERG

R87572

2.9

0.4

cyclin G2*

CCNG2

H18981

2.0

0.8

Fibronectin (fn precursor)

FN1

W57892

2.1

1.4

Dermal fibroblast elastin‡

WDR1

R49766

2.2

0.6

Transporter/signal transduction

Transcription factors/cell cycle

Cellular component/cytoskeleton

Other/unknown Williams-Beuren syndrome chromosome region 5‡

WBSCR5

AF086239

2.1

0.6

kiaa1766 protein

KIAA1766

AI800340

2.1

1.6

kiaa0125 gene product‡

KIAA0125

H65343

2.2

0.6

Melanoma-associated gene‡

D2S448

R09509

2.0

0.5

lim protein slimmer*

FHL1

AA134667

2.3

1.1

kiaa0581 protein‡

PLCB1

AA011027

2.2

0.8

ralgds-like gene

RGL

AA013091

2.1

1.9

Cysteine and tyrosine-rich 1*

CYYR1

AA125921

2.0

0.8

Hypothetical protein flj14054‡

FLJ14054

H68441

2.9

0.8

AA042839

2.3

0.8

cDNA dkfzp586e1624* Acyl-protein thioesterase

LYPLA2

R38687

2.3

3.3

Hypothetical protein flj10628‡

FLJ10628

AA400002

2.3

0.7

The Log2 ratios of SDF/FDF were the average values of 6 hybridizations in 3 experiments. Error estimate is given as SD. Probability of differential gene expression in relation to all high- quality signals was determined with the Student t test. MHC indicates major histocompatibility complex. Accession numbers from GenBank are provided. *P ⬍ .01. †P ⬍ .0001. ‡P ⬍ .001.

believed to play a role in neuronal differentiation. The ESAM1, endomucin 2, and galectin-1 are adhesion proteins that might have an impact on the interaction of CD34⫹/CD38⫺ cells with the microenvironment. In contrast, the CD34⫹/CD38⫹ population displayed higher expression of several genes that are associated with lineage commitment for mature blood cells, as myeoloperoxidase, T-cell receptor, mast cell carboxypeptidase a, and the erythroid transcription factors EKLF and GATA1. LEF1 is a nuclear protein expressed in pre-B and T cells that can form complexes with ␤ catenin and is part of the WNT pathway.41 Other genes, such as cyclin B2, cyclin PCNA, and ki-67, might influence cell cycle kinetics in the CD34⫹/CD38⫹ population.42 Higher expression of ki-67 in the CD34⫹/CD38⫹ fraction and a correlation between CD38 expression and cell cycle status have been reported by other authors.43,44

Subsequently the gene expression profile of SDF versus FDF was analyzed. The SDF is highly enriched in functionally primitive and asymmetrically dividing cells, whereas cells from the FDF give rise to more committed progenitors. Indeed, we found several markers that have been associated with HSCs to be highly expressed in the SDF. Prominin (CD133), a marker associated with mobility and primitive function, was among the highest differentially expressed genes (13-fold).45,46 The multiple drug resistance gene 1 (MDR1) and the complement component 1 receptor 1 (clqr1), reported to be characteristic for stem cells, were both increased 5-fold.47 Several transcription factors that might have an impact on the hematopoietic differentiation were also higher expressed in the SDF, for example, the homeodomain proteins HOXA9 (3.9-fold), cdx1, and hesx1. The Kruppel-like factor 12 (klf12) has been reported to be highly expressed in the hematopoietic system of zebrafish.48 ERG



WAGNER et al

Table 4. Genes down-regulated more than 4-fold in the SDF versus FDF Gene

Symbol

Accession no.

Ratio (Log2)

SD

Cell surface receptors/ligands CD36 antigen*

CD36

R09416

⫺2.3

0.6

Cadherin 1 type l/e-cadherin (epithelial)†

CDH1

R89322

⫺3.6

0.5

Transporter/signal transduction xk mRNA for membrane transport protein‡

XK

H18932

⫺2.3

1.1

(pwd) gene†

ATP7B

H84807

⫺2.3

0.3

KLF1

H65733

⫺2.7

0.7

␤-Globin (hbb)‡

HBB

N69245

⫺2.6

1.1

␥-Globin*

HBG1

H53258

⫺2.4

0.6

␥-Globin

HBG1

H93108

⫺2.2

1.5

Hemoglobin ␧-1‡

HBE1

H79533

⫺2.1

0.9

Adducin ␤ subunit 63 kDa isoform*

ADD2

R19254

⫺2.9

1.0

N-acetylglucosamine-6-o-sulfotransferase (glcnac6st)‡

CHST2

N80041

⫺2.3

0.9

Cystathionine-␤-synthase (cbs)‡

CBS

N54505

⫺2.0

1.1

5-Aminolevulinate synthase 2 (alas2)‡

ALAS2

R98713

⫺2.0

1.3

Follistatin‡

FST

R01927

⫺2.4

1.3

AI589889

⫺2.1

1.4

H95237

⫺2.2

0.9

H93462

⫺2.5

0.8

R18099

⫺2.4

0.6

Transcription factors/cell cycle Erythroid Kruppel-like factor EKLF gene* Cellular component/cytoskeleton

Other/unknown

EST similar to lphuc1 apolipoprotein c-i precursor Precursor of apolipoprotein ci (apo ci)‡

APOC1

Similar to ae-binding protein 2* gs3955*

GS3955

The Log2 ratios of SDF/FDF were the average values of 6 hybridizations in 3 experiments. Error estimate is given as SD. Probability of differential gene expression in relation to all high-quality signals was determined with the Student t test. Accession numbers from GenBank are provided. *P ⬍ .001. †P ⬍ .0001. ‡P ⬍ .01.

resembles a transcription factor that is involved in a form of acute myeloic leukemia by chromosomal translocation49; IFI16 may function as a transcriptional repressor that is present in myeloid precursors (CD34⫹) and throughout monocyte development, but its expression is down-regulated in erythroid and polymorphonuclear precursor cells.50,51 Several adhesion proteins, including platelet endothelial cell adhesion molecule 1 (PECAM-1), protocadherin ␤4, intercellular adhesion molecule 3 (ICAM-3), leukocyte factor antigen 1 (LFA-1), and integrin ␤2, were also more highly expressed in the SDF and might be involved in the homing of stem cells or their interaction with the microenvironment. Selectin L (SELL) is a cell surface component that functions in leukocyte-endothelial cell interactions and was reported to be increas-

ingly expressed on CD34⫹ CB cells during gestation.52 The cyclin G2 is a negative regulator of cell cycle progression as observed in murine B cells responding to growth inhibitory stimuli and may thus have an impact on the low proliferation of the SDF.53 HLA genes are organized in chromosome 6 under the additional transcriptional control of a locus control region (LCR).54 Thirteen independent HLA genes were highly up-regulated in the SDF. Although HLA-DR is usually considered to be expressed on more mature cells, it has been reported that CD34⫹/CD38⫺/ HLA-DR⫹ cells have the potential to give rise to all hematopoietic lineages.2,4 We have shown that the CD34⫹/CD38⫺/HLA-DR⫹ subset from fetal tissues contained a higher concentration of candidate stem cells.4

Table 5. Genes with higher expression in differentially more primitive population CD34ⴙCD38ⴚ Gene

Symbol

Accession no.

SDF

Ratio

SD

Ratio

SD

Osteopontin

SPP1

R59298

1.0

0.4

3.0

1.8

Phospholipase c-like 2

PLCL2

W72046

0.9

0.1

1.1

0.7

kiaa0337 gene

KIAA0337

AA236696

1.0

0.2

2.0

0.7

sh3 binding protein

SH3BP5

N52513

1.1

0.2

1.5

1.2

Multidrug resistance gene 1 (MDR1)

ABCB1

AA135957

0.8

0.3

2.1

0.5

mn1

MN1

R59212

0.9

0.2

1.1

0.9

Prepromultimerin

MMRN

AA423867

0.9

0.3

2.1

1.4

Prostate androgen-induced RNA

TMEPAI

H14511

0.8

0.2

1.1

0.6

rbp-ms/type 3

RBPMS

AA045722

1.3

0.4

1.4

1.1

Histones h2b.1 and h2a

H2BFQ

R98471

1.1

0.4

1.0

0.4

Hypothetical protein dkfzp761d221

DKFZP761D221

AA018443

1.0

0.2

1.2

1.2

Ecotropic viral integration site 2a

EVI2A

AI701091

0.9

0.5

1.1

0.7

Hypothetical protein smap31

SMAP31

AA504137

1.2

0.3

1.7

1.4

Hypothetical protein pp1044

PP1044

N34427

0.9

0.1

1.6

1.2

PR-domain zinc finger protein 1

PRDM1

AI692550

1.0

0.3

1.1

1.0

Accession numbers are from GenBank.



683

Table 6. Genes with higher expression in committed progenitors CD34ⴙCD38ⴙ Gene

Symbol

Accession no.

FDF

Ratio, log2

SD

Ratio, log2

SD

CD36 antigene

CD36

R09416

⫺1.2

0.4

⫺2.3

0.6

Gs3955

GS3955

R18099

⫺1.2

0.3

⫺2.4

0.6

N-acetylglucosamine-6-o-sulfotransferase

CHST2

N80041

⫺1.1

0.3

⫺2.3

0.9

Chromosome 20 open reading frame 129

C20ORF129

N62469

⫺1.4

0.3

⫺1.0

0.2

Solute carrier family 11

SLC11A3

T97261

⫺1.2

0.5

⫺1.0

0.5

L-histidine decarboxylase

HDC

AI695885

⫺3.7

0.7

⫺1.8

1.1

Erythroid Kruppel-like factor EKLF

KLF1

H65733

⫺1.6

0.3

⫺2.7

0.7

kiaa0473 protein

DNAJC6

N38951

⫺1.1

0.2

⫺1.6

0.6

Lymphoid enhancer-binding factor-1

LEF1

W81128

⫺1.6

0.7

⫺1.1

0.5

kiaa0124 gene

BOP1

AA482401

⫺0.9

0.7

⫺1.0

0.6

Kell blood group protein

KEL

H61054

⫺1.1

0.8

⫺1.7

1.4

Transcription factor eryfl

GATA1

R06446

⫺0.9

0.4

⫺1.2

0.4

In the FDF genes of the ␤-globin locus showed a much higher expression. The erythroid transcription factors gata1 and EKLF are well known regulators of the corresponding LCR and were more highly expressed in this population. Furthermore, other genes of the erythroid lineage are more highly expressed in the FDF: ALAS2, involved in heme biosynthesis; glycophorin b, a major sialoglycoprotein of the human erythrocyte membrane; and ecadherin, a calcium-dependent adhesion protein interacting preferentially in a homophilic manner that is restricted to the erythroid lineage in human bone marrow.55,56 Thus the observed differential gene expression in the SDF and FDF is compatible with the observation that slow and asymmetrically dividing cells indeed represent a functionally more primitive population. On the other hand, differential gene expression in the FDF might imply that this fraction differentiated into several hematopoietic lineages during the 7 days of cultivation. A shorter incubation time might help to further

elucidate the molecular and genetic differences between cells that self-renew versus cells that are destined to differentiate. Analysis of cell populations that differ by only one or 2 divisions might help to identify the most salient molecular differences and are underway. Several studies on the gene expression of murine and human HSC fractions have been published recently.9-15 A combination of different data sets and the determination of the overlap in differential gene expression may shed light on gene products that are important for stem cell function.57-59 In comparison with 3 different microarray approaches9,10,15 we estimated the overlap of genes in different fractions of murine and human HSCs. Among the genes that were up-regulated in several fractions were frizzled 6 (FZD6) that might function as a receptor for the Wnt-pathway,60 RNA-binding protein with multiple splicing (RBPMS), MDR1, JAK3, and the homeodomain protein HOXA9. HOXA9 knockout mice showed defects in hematopoiesis,

Table 7. Comparative analysis with 3 different approaches

Researcher

Stem cell population/ baseline cell population

RNA amplification

Microarray, no. of ESTs characterized/uncharacterized

Overlap of HSCenriched genes with our CD34ⴙCD38ⴚ-enriched genes

Own data

HCB: CD34⫹CD38⫺/

In vitro

cDNA: Human Transcriptome

27

355

5 (RBPMS, R-RAS,

38 (HOXA9, RBPMS, FZD6, JAK3,

CD34⫹CD38⫹ HCB: SDF/FDF Ivanova et al9

HFL: CD34⫹CD38⫺/Lin⫹ MBM: Lin⫺c-Kit⫹-Scal⫹ Rholow/Lin⫹

transcription

Overlap of HSC-enriched genes with our SDf-enriched genes

Microarray, ⬎15 000/51 145

(2 rounds) In vitro

Oligonucleotide microarrays

transcription

(Affymetrix) HG-U95 A, B, C,

PRDM1, EMCN-

PRDM1, CYSLT1, P2RY5,

(2 rounds)

D, E: ⬎10 000/60 000 MG-

PENDING, CASP12 )

DNMT3A, UTRN, NFKB1A, CIQR1,

MFL: Lin⫺AA4.1⫹ c-Kit⫹

U74v2 A, B, C: ⬎6000/36 000

CHN2, CPXM1, LMYC1, MYCN,

Scal⫹/Lin⫹

EZNF, TC10, KCNMB1, SV2, FZD8, ITGAV, MMP7, NCAM2, ADA, KLF12, SOX4, LANO, FHL1, MANBA, SSBP2, MAFB, JM1, ZNF43, ETV6, CLECSF2, SIGIRRPENDING, H2-DMA, GIG1)

Ramalho-

MBM: Hoechst33342low

In vitro

Oligonucleotide microarrays

5 (RBPMS, R-RAS,

15 (HOXA9, RBPMS, MDR1, JAK3,

Santos

Scal⫹ c-Kit⫹

transcription

(Affymetrix) U74Av2 A: ⬎

HSP40, NCKAP1,

ITM2, GNG11, TC10, LTB, MYCN,

et al10

CD34⫺Lin⫺/Lin⫹

(2 rounds)

6000/12 000

CASP12)

STXBP3, ITM3, NAALADASEL,

MBM: c-Kit⫹ Thy1.1low

SMART-PCR

GALNT3, HLA-DMA, CKIP-1) Terskikh et al15

Scal⫹Lin⫺/average

(19 cycles)

cDNA Atlas1.2 (Clontech) 1200

0

5 (HOXA9, FZD6, MDR1, ITGB2, PCLG2)

expression level of 7 populations Mouse-human homolog genes were determined as described in “Materials and methods.” HCB indicates human cord blood; HFL, human fetal liver; MBM, mouse bone marrow; SDF, slow-dividing fraction within CD34⫹CD38⫺ cell population; FDF, fast-dividing fraction in CD34⫹CD38⫺ cell population; MFL, mouse fetal liver.


WAGNER et al


Figure 3. Confirmation of differential expression by real-time RT-PCR. Semiquantitative RT-PCR by LightCycler analysis was used to verify differential gene expression of 7 genes in FDF and SDF in relation to 18s rRNA. HLA-A and CD133 were higher expressed in the SDF. ␥-Globin, CD36, and cadherin-e (not shown) were more highly expressed in the FDF. GAPDH and ubiquitin did not show significant differential expression and were used as additional controls. All figures belong to the same run of 1 RNA sample. A high correlation between RT-PCR and microarray analysis was observed. Error bars represent SD of at least 3 experiments.

whereas HOXA9 overexpression increased transplantable lymphoid-myeloid long-term repopulating cells.61,62 A recent cDNA microarray analysis identified several target genes of HOXA9 that may mediate the biologic effects of this protein.63 Jak3, a nonreceptor tyrosine kinase, might be a regulator for differentiation because defects in Jak3 led to severe combined immunodeficiency.64,65 Although the different microarray studies compared used different starting material and methods this comparative analysis has revealed several candidate genes that were predominantly expressed in primitive subsets of HSCs. It has been demonstrated before that primary human CD34⫹ cells from fetal liver, umbilical CB, and adult bone marrow and peripheral blood could form various types of pseudopodia morphologies.66,67 Lamellipodia-type processes are associated with migration.67 An unexpected observation in this study is that the majority

of cells in the SDF demonstrated an elongated cell shape in comparison to the round cell shape in the FDF. Flow cytometric analysis has confirmed the significantly higher expression of CD133. CD133 has previously been described to be expressed on plasma membrane protrusions.68,69 Observation using immunefluorescence microscopy and time-lapse camera system confirmed that cells from the SDF possessed lamellipodia with strong expression of CD133. Evidence indicates that the Rh-123low progenitor cell population is enriched in cells that are able to reconstitute hematopoiesis permanently.70 We have demonstrated that the SDF is highly enriched in Rh-123low cells as compared to the FDF. The MDR1 efflux pump has been reported to be responsible for eliminating Rh-123 and is thus associated with rhodamine low phenotype and with long-term repopulating activity.71 This is also in line with our

Figure 4. SDF has fibroblastoid morphology. The SDF (A-B) and the FDF (C-D) were separated according to their PKH26 staining after 1 week. Cells in the SDF are on average smaller; 69% of them revealed membrane protrusions. In contrast, cells in the FDF had a round morphology and only 16% displayed podia formation. CD133 (prominin) is more strongly expressed in the SDF and accumulates at the tip of lamellipodia (E-F; G-H; red ⫽ PKH26; green ⫽ CD133). Error bars represent SD. Cells were observed with a 40⫻ immersion objective (40⫻ HI; Olympus Optical, Hamburg, Germany), microscope model IX-70 (Olympus Optical) connected to a video camera (Colorview XS; Soft Images Systems, Mu¨ nster, Germany; resolution: pixel size 6.7 ⫻ 6.7 ␮m; NA of 40 ⫻ objective: 1.0).



685

SDF of CD34⫹/CD38⫺ population. The molecular characteristics provided further evidence that this cell fraction is associated with primitive function and asymmetric division. The data obtained in this study will serve as a basis for future analyses, in particular of genes with yet unknown function that might play a significant role in asymmetric division and self-renewal.

Figure 5. SDF can efflux Rh-123. The vast majority of the SDFs (PKH26⫹) can efflux Rh-123, whereas the FDFs (PKH26⫺) were stained as Rh-123high (A). With the inclusion of the P-glycoprotein inhibitor cyclosporin (CsA) efflux of Rh-123 in the SDF was effectively blocked (B; representative for 3 experiments).

observation that the MDR1 gene is about 5-fold higher expressed in the SDF. In summary, we have used the novel Human Transcriptome Microarray tool to determine the gene expression profile of the

Acknowledgments We thank Bernhard Korn and the Resource Center and Primary Database (RZPD) for the supply of the IMAGE clones and their sequence verification as described in the publication on the production of the Human Transcriptome Microarray, Eike Buss for the assistance with the rhodamine-efflux assay, Heidi Rossmann for the assistance with the LightCycler RT-PCR, and Michael Punzel for valuable advice to these studies.

References 1. Ho AD, Punzel M. Hematopoietic stem cells: can old cells learn new tricks? J Leukoc Biol. 2003;73: 547-555.

14. Park IK, He Y, Lin F, et al. Differential gene expression profiling of adult murine hematopoietic stem cells. Blood. 2002;99:488-498.

2. Huang S, Terstappen LW. Lymphoid and myeloid differentiation of single human CD34⫹, HLA-DR⫹, CD38⫺ hematopoietic stem cells. Blood. 1994;83: 1515-1526.

15. Terskikh AV, Miyamoto T, Chang C, Diatchenko L, Weissman IL. Gene expression analysis of purified hematopoietic stem cells and committed progenitors. Blood. 2003;102:94-101.

3. Ishikawa F, Livingston AG, Minamiguchi H, Wingard JR, Ogawa M. Human cord blood long-term engrafting cells are CD34⫹ CD38⫺. Leukemia. 2003;17:960-964.

16. Huang S, Law P, Francis K, Palsson BO, Ho AD. Symmetry of initial cell divisions among primitive hematopoietic progenitors is independent of ontogenic age and regulatory molecules. Blood. 1999; 94:2595-2604.

4. Huang S, Law P, Young D, Ho AD. Candidate hematopoietic stem cells from fetal tissues, umbilical cord blood versus adult bone marrow and mobilized peripheral blood. Exp Hematol. 1998;26: 1162-1171. 5. Miller JS, McCullar V, Punzel M, Lemischka IR, Moore KA. Single adult human CD34(⫹)/Lin(⫺)CD38(⫺) progenitors give rise to natural killer cells, B-lineage cells, dendritic cells, and myeloid cells. Blood. 1999;93:96-106. 6. Liu H, Verfaillie CM. Myeloid-lymphoid initiating cells (ML-IC) are highly enriched in the rhodamine-c-kit(⫹)CD33(⫺)CD38(⫺) fraction of umbilical cord CD34(⫹) cells. Exp Hematol. 2002;30: 582-589. 7. Morrison SJ, Weissman IL. The long-term repopulating subset of hematopoietic stem cells is deterministic and isolatable by phenotype. Immunity. 1994;1:661-673 8. Berardi AC, Wang A, Levine JD, Lopez P, Scadden DT. Functional isolation and characterization of human hematopoietic stem cells. Science. 1995;267:104-108. 9. Ivanova NB, Dimos JT, Schaniel C, Hackney JA, Moore KA, Lemischka IR. A stem cell molecular signature. Science. 2002;298:601-604. 10. Ramalho-Santos M, Yoon S, Matsuzaki Y, Mulligan RC, Melton DA. “Stemness”: transcriptional profiling of embryonic and adult stem cells. Science. 2002;298:597-600. 11. Phillips RL, Ernst RE, Brunk B, et al. The genetic program of hematopoietic stem cells. Science. 2000;288:1635-1640. 12. Gomes I, Sharma TT, Edassery S, Fulton N, Mar BG, Westbrook CA. Novel transcription factors in human CD34 antigen-positive hematopoietic cells. Blood. 2002;100:107-119. 13. Gomes I, Sharma TT, Mahmud N, et al. Highly abundant genes in the transcriptosome of human and baboon CD34 antigen-positive bone marrow cells. Blood. 2001;98:93-99.

17. Punzel M, Wissink SD, Miller JS, Moore KA, Lemischka IR, Verfaillie CM. The myeloid-lymphoid initiating cell (ML-IC) assay assesses the fate of multipotent human progenitors in vitro. Blood. 1999;93:3750-3756. 18. Punzel M, Zhang T, Liu D, Eckstein V, Ho AD. Functional analysis of initial cell divisions defines the subsequent fate of individual human CD34(⫹)CD38(⫺) cells. Exp Hematol. 2002;30: 464-472. 19. Mahmud N, Devine SM, Weller KP, et al. The relative quiescence of hematopoietic stem cells in nonhuman primates. Blood. 2001;97:3061-3068. 20. Cheng T, Rodrigues N, Shen H, et al. Hematopoietic stem cell quiescence maintained by p21cip1/ waf1. Science. 2000;287:1804-1808. 21. Reems JA, Torok-Storb B. Cell cycle and functional differences between CD34⫹/CD38hi and CD34⫹/38lo human marrow cells after in vitro cytokine exposure. Blood. 1995;85:1480-1487. 22. Gothot A, Pyatt R, McMahel J, Rice S, Srour EF. Assessment of proliferative and colony-forming capacity after successive in vitro divisions of single human CD34⫹ cells initially isolated in G0. Exp Hematol. 1998;26:562-570. 23. Gothot A, van der Loo JC, Clapp DW, Srour EF. Cell cycle-related changes in repopulating capacity of human mobilized peripheral blood CD34(⫹) cells in non-obese diabetic/severe combined immune-deficient mice. Blood. 1998;92:2641-2649. 24. Horan PK, Melnicoff MJ, Jensen BD, Slezak SE. Fluorescent cell labeling for in vivo and in vitro cell tracking. Methods Cell Biol. 1990;33: 469-490. 25. Lee GM, Fong SS, Oh DJ, Francis K, Palsson BO. Characterization and efficacy of PKH26 as a probe to study the replication history of the human hematopoietic KG1a progenitor cell line. In Vitro Cell Dev Biol Anim. 2002;38:90-96. 26. Yan F, Collector MI, Tyszko S, Sharkis SJ. Using

divisional history to measure hematopoietic stem cell self-renewal and differentiation. Exp Hematol. 2003;31:56-64. 27. Richter A, Schwager C, Hentze S, Ansorge W, Hentze MW, Muckenthaler M. Comparison of fluorescent tag DNA labeling methods used for expression analysis by DNA microarrays. Biotechniques. 2002;33:620-630. 28. Muckenthaler M, Richter A, Gunkel N, et al. Relationships and distinctions in iron-regulatory networks responding to interrelated signals. Blood. 2003;101:3690-3698. 29. Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001;98:5116-5121. 30. Brazma A, Hingamp P, Quackenbush J, et al. Minimum information about a microarray experiment (MIAME)—toward standards for microarray data. Nat Genet. 2001;29:365-371. 31. Rossmann H, Bachmann O, Vieillard-Baron D, Gregor M, Seidler U. Na⫹/HCO3⫺ cotransport and expression of NBC1 and NBC2 in rabbit gastric parietal and mucous cells. Gastroenterology. 1999;116:1389-1398. 32. Fruehauf S, Breems DA, Knaan-Shanzer S, et al. Frequency analysis of multidrug resistance-1 gene transfer into human primitive hematopoietic progenitor cells using the cobblestone area-forming cell assay and detection of vector-mediated P-glycoprotein expression by rhodamine-123. Hum Gene Ther. 1996;7:1219-1231. 33. Schiedlmeier B, Kuhlcke K, Eckert HG, Baum C, Zeller WJ, Fruehauf S. Quantitative assessment of retroviral transfer of the human multidrug resistance 1 gene to human mobilized peripheral blood progenitor cells engrafted in nonobese diabetic/severe combined immunodeficient mice. Blood. 2000;95:1237-1248. 34. Punzel M, Liu D, Zhang T, Eckstein V, Miesala K, Ho AD. The symmetry of initial divisions of human hematopoietic progenitors is altered only by the cellular microenvironment. Exp Hematol. 2003; 31:339-347. 35. Hubbard T, Barker D, Cameron G, et al. The Ensembl genome database project. Nucleic Acids Res. 2002;30:38-41. 36. Radelof U, Maurer J, Korn B. Human UnigeneSet-RZPD3—ein Klonsatz fu¨ r das gesamte Genom. Transkript. 2002;11:67. 37. Abbott A. Gene centre chips in with better route to microarrays. Nature. 2002;420:3.



WAGNER et al

38. Schwartz GN, Warren MK, Sakano K, et al. Comparative effects of insulin-like growth factor II (IGF-II) and IGF-II mutants specific for IGF-II/ CIM6-P or IGF-I receptors on in vitro hematopoiesis. Stem Cells.;14:337-350. 39. Chaudhary PM, Roninson IB. Expression and activity of P-glycoprotein, a multidrug efflux pump, in human hematopoietic stem cells. Cell. 1991;66:85-94. 40. Buijs A, Sherr S, van Baal S, et al. Translocation (12;22) (p13;q11) in myeloproliferative disorders results in fusion of the ETS-like TEL gene on 12p13 to the MN1 gene on 22q11. Oncogene. 1995;10:1511-1519. 41. Reya T, O’Riordan M, Okamura R, et al. Wnt signaling regulates B lymphocyte proliferation through a LEF-1 dependent mechanism. Immunity. 2000;13:15-24. 42. Schluter C, Duchrow M, Wohlenberg C, et al. The cell proliferation-associated antigen of antibody Ki-67: a very large, ubiquitous nuclear protein with numerous repeated elements, representing a new kind of cell cycle-maintaining proteins. J Cell Biol. 1993;123:513-522. 43. Jordan CT, Yamasaki G, Minamoto D. High-resolution cell cycle analysis of defined phenotypic subsets within primitive human hematopoietic cell populations. Exp Hematol. 1996;24:1347-1355. 44. Hao QL, Shah AJ, Thiemann FT, Smogorzewska EM, Crooks GM. A functional comparison of CD34⫹ CD38⫺ cells in cord blood and bone marrow. Blood. 1995;86:3745-3753. 45. Yin AH, Miraglia S, Zanjani ED, et al. AC133, a novel marker for human hematopoietic stem and progenitor cells. Blood. 1997;90:5002-5012. 46. Miraglia S, Godfrey W, Yin AH, et al. A novel fivetransmembrane hematopoietic stem cell antigen: isolation, characterization, and molecular cloning. Blood. 1997;90:5013-5021. 47. Danet GH, Luongo JL, Butler G, et al. C1qRp defines a new human stem cell population with hematopoietic and hepatic potential. Proc Natl Acad Sci U S A. 2002;99:10441-10445. 48. Oates AC, Pratt SJ, Vail B, et al. The zebrafish klf gene family. Blood. 2001;98:1792-1801. 49. Kong XT, Ida K, Ichikawa H, et al. Consistent detection of TLS/FUS-ERG chimeric transcripts in acute myeloid leukemia with t(16;21)(p11;q22) and identification of a novel transcript. Blood. 1997;90:1192-1199.

50. Dawson MJ, Elwood NJ, Johnstone RW, Trapani JA. The IFN-inducible nucleoprotein IFI 16 is expressed in cells of the monocyte lineage, but is rapidly and markedly down-regulated in other myeloid precursor populations. J Leukoc Biol. 1998; 64:546-554. 51. Briggs RC, Briggs JA, Ozer J, et al. The human myeloid cell nuclear differentiation antigen gene is one of at least two related interferon-inducible genes located on chromosome 1q that are expressed specifically in hematopoietic cells. Blood. 1994;83:2153-2162. 52. Surbek DV, Steinmann C, Burk M, Hahn S, Tichelli A, Holzgreve W. Developmental changes in adhesion molecule expressions in umbilical cord blood CD34 hematopoietic progenitor and stem cells. Am J Obstet Gynecol. 2000;183:1152-1157. 53. Horne MC, Donaldson KL, Goolsby GL, et al. Cyclin G2 is up-regulated during growth inhibition and B cell antigen receptor-mediated cell cycle arrest. J Biol Chem. 1997;272:12650-12661. 54. Li Q, Peterson KR, Fang X, Stamatoyannopoulos G. Locus control regions. Blood. 2002;100:30773086. 55. Armeanu S, Buhring HJ, Reuss-Borst M, Muller CA, Klein G. E-cadherin is functionally involved in the maturation of the erythroid lineage. J Cell Biol. 1995;131:243-249. 56. Buhring HJ, Muller T, Herbst R, et al. The adhesion molecule E-cadherin and a surface antigen recognized by the antibody 9C4 are selectively expressed on erythroid cells of defined maturational stages. Leukemia. 1996;10:106-116. 57. Fortunel NO, Otu HH, Ng HH, et al. Comment on “stemness: transciptional profiling of embryonic and adult stem cells” and “a stem cell molecular signature.” Science. 2003;302:393. 58. Ivanova NB, Dimos JT, Schaniel C, et al. Response to comments on “stemness”: transciptional profiling of embryonic and adult stem cells and “a stem cell molecular signature.” Science. 2003;302:393.

61. Lawrence HJ, Helgason CD, Sauvageau G, et al. Mice bearing a targeted interruption of the homeobox gene HOXA9 have defects in myeloid, erythroid, and lymphoid hematopoiesis. Blood. 1997;89:1922-1930. 62. Thorsteinsdottir U, Mamo A, Kroon E, et al. Overexpression of the myeloid leukemia-associated Hoxa9 gene in bone marrow cells induces stem cell expansion. Blood. 2002;99:121-129. 63. Dorsam ST, Ferrell CM, Dorsam GP, et al. The transcriptome of the leukemogenic homeoprotein HOXA9 in human hematopoietic cells. Blood. 2004;103:1676-1684. 64. Candotti F, Oakes SA, Johnston JA, et al. Structural and functional basis for JAK3-deficient severe combined immunodeficiency. Blood. 1997; 90:3996-4003. 65. Roberts JL, Lengi A, Brown SM, et al. Janus kinase 3 (JAK3) deficiency: clinical, immunologic and molecular analyses of 10 patients and outcomes of stem cell transplantation. Blood. 2004; 103:2009-2018. 66. Francis K, Ramakrishna R, Holloway W, Palsson BO. Two new pseudopod morphologies displayed by the human hematopoietic KG1a progenitor cell line and by primary human CD34(⫹) cells. Blood. 1998;92:3616-3623. 67. Fruehauf S, Srbic K, Seggewiss R, Topaly J, Ho AD. Functional characterization of podia formation in normal and malignant hematopoietic cells. J Leukoc Biol. 2002;71:425-432. 68. Corbeil D, Roper K, Hellwig A, et al. The human AC133 hematopoietic stem cell antigen is also expressed in epithelial cells and targeted to plasma membrane protrusions. J Biol Chem. 2000;275:5512-5520. 69. Corbeil D, Roper K, Fargeas CA, Joester A, Huttner WB. Prominin: a story of cholesterol, plasma membrane protrusions and human pathology. Traffic. 2001;2:82-91.

59. Li L, Akashi K. Unraveling the molecular components and genetic blueprints of stem cells. Biotechniques. 2004;35:1233-1239.

70. Bertoncello I, Hodgson GS, Bradley TR. Multiparameter analysis of transplantable hemopoietic stem cells, I: the separation and enrichment of stem cells homing to marrow and spleen on the basis of rhodamine-123 fluorescence. Exp Hematol. 1985;13:999-1006.

60. Walsh J, Andrews PW. Expression of Wnt and Notch pathway genes in a pluripotent human embryonal carcinoma cell line and embryonic stem cell. APMIS. 2003;111:197-210.

71. Chaudhary PM, Roninson IB. Expression and activity of P-glycoprotein, a multidrug efflux pump, in human hematopoietic stem cells. Cell. 1991;66:85-94.