Gene expression analysis of purified hematopoietic stem cells and committed progenitors

Alexey V. Terskikh, Toshihiro Miyamoto, Cynthia Chang, Luda Diatchenko, Irving L. Weissman


Lifelong self-renewal is a unique property of somatic stem cells. Recently, several primitive multipotent yet committed (non—self-renewing) hematopoietic progenitor populations were identified in mouse bone marrow. We have characterized the expression of 1200 selected mouse genes using the Atlas cDNA array in highly purified hematopoietic stem cells (HSCs) and 6 closely related progenitor populations: common myeloid progenitors (CMPs), granulocyte-macrophage progenitors (GMPs), megakaryocyte-erythrocyte progenitors (MEPs), common lymphoid progenitors (CLPs), and pro-T and pro-B cells. Cluster analysis revealed that nearly half of all differentially expressed transcripts are associated with HSCs, supporting the notion of an active transcriptional status of HSCs. Genes found enriched in the HSC cluster encompass many developmentally regulated genes, some previously associated with HSC self-renewal. In contrast, genes that are enriched in committed progenitors are mostly associated with hematopoietic differentiation, immune regulation, and metabolism. Thus, the transition from HSCs toward committed progenitors correlates with the down-regulation of a large number of HSC-associated genes and progressive up-regulation of a limited number of lineage-specific genes. These genetic analyses revealed both quantitative and qualitative differences between the transcripts associated with HSCs versus downstream progenitors and produced a list of the candidate genes, potentially involved in HSC self-renewal. (Blood. 2003;102:94-101)


Somatic stem cells are unique among all other cells in the body in respect to their lifelong self-renewal capacity (reviewed in Weissman1; Gage2; and Watt and Hogan3). This proliferation is tightly regulated to maintain a fine balance (homeostasis) between the number of stem cells and their differentiated progeny. Hematopoietic stem cells (HSCs) have been extensively characterized in the last decade and represent a rare case wherein a highly purified and apparently functionally homogeneous population of cells can be prospectively isolated by flow cytometry.4,5 This feature of the hematopoietic system was the basis for previous genetic analyses of primary mouse HSCs.6,7 However, in both studies, HSCs were compared with the rest of the hematolymphoid population of bone marrow and fetal liver cells to the entire non—stem cell compartment. This approach can overlook fine differences in gene expression between HSCs and their immediate committed progeny, because of a very small number of progenitor cells in bone marrow and in fetal liver.8,9 However, this fine difference is of great interest. Indeed, it is between HSCs and their closest committed progeny that lies the apparently invisible borderline of self-renewal capacity.

Recently, a number of committed progenitor populations have been identified in mouse bone marrow and prospectively isolated by flow cytometry.8,9 As of today, these populations represent the closest HSC descendants that can be prospectively isolated by flow cytometry and are still oligopotent (can give rise to some, but not all lineage types), yet have already lost the capacity for self-renewal.8,9 Little is known about the molecular mechanisms underlying the biology of HSCs and committed progenitors. The expression of several transcription factors known to play roles in hematopoiesis was recently examined in these populations using reverse transcription—polymerase chain reaction (RTPCR) analysis.9 The complexity of molecular mechanisms underlying differentiation and commitment was recently underscored by an induced cell-fate conversion of committed lymphoid progenitors (CLPs) to myelomonocytic fates.10 Recently, we have demonstrated that prospectively purified single common myeloid progenitors (CMPs) coexpress myeloerythroid but not lymphoid genes, whereas single CLPs coexpress T and B lymphoid but not myeloid genes.11 Importantly, we have demonstrated that on the functional level, within the limits of our in vitro and in vivo assays, these cells are pure by phenotype, function, and gene expression profile. More than 90% of megakaryocyte-erythrocyte progenitors (MEPs) assayed at the single-cell level by RT-PCR expressed erythropoietin receptor (EpoR) and GATA-1, but not myeloperoxidase (MPO) or granulocyte colony-stimulating factor (GCSF) receptor, and more than 90% of a single granulocyte-macrophage progenitor (GMP) expressed the reverse profile (MPO+, GCSF-R+, EpoR-, GATA-1-).

In the present study we took advantage of cDNA array technology to investigate systematically the expression profile of 1200 known mouse genes in highly purified HSCs and progenitor populations, namely CMPs, granulocyte-macrophage progenitors (GMPs), megakaryocyte-erythrocyte progenitors (MEPs), CLPs, pro-T cells, and pro-B cells (Figure 1). Because of the small initial number of cells available for the analysis, a SMART-PCR amplification technology was applied. This approach was recently demonstrated to be highly reliable for gene expression array profiling from minute amounts of total RNA.12

Figure 1.

The experimental design. The surface markers used for FACS isolation are shown above the corresponding cells. The lineage relationship among the stem and progenitor populations is shown by the arrows. The functional in vitro and in vivo assays confirming the lineage relationship for each population were described previously.5,8,9

Materials and methods

Cell isolation

All cell populations were isolated from C57BL/Ka-Thy1 by flow cytometry (Vantage SE; Becton Dickinson, San Jose, CA) as previously described.5,8,9 Briefly, HSCs and various progenitor populations were prospectively isolated using a highly modified triple-laser fluorescence activated cell sorter (FACS Vantage). The 5-color sort using both positive and negative gates in multiple channels usually gives rise to cells of more than 99% phenotypic purity, avoiding cosorting cells stained in a nonspecific manner. The sorted cells were subjected to 2 additional rounds of sorting using the same gate to eliminate contaminating cells and doublets. The sample line was washed out by 75% ethanol followed by saline between each round of sorting to eliminate remaining cells. About 400 cells of the final, triple-sorted population (data presented in Figure 2) were analyzed to find no cells outside of the sorting gates defined for each population.

Figure 2.

The actual phenotype and purity of each isolated population was confirmed by the reanalysis after the third sorting procedure. The most informative dot plots (5% probability) with outliners are shown for each population; previous sorting gates are indicated above. About 400 cells were collected for each population, representing approximately 10% of the total number of cells used for RNA isolation. No cells were found outside of the sorting gates specified for each population.

RNA isolation and first-strand cDNA strand synthesis

About 4000 cells from each population were lysed using Trizol, and total RNA was purified according to the manufacturer's instructions (Life Technologies, Rockville, MD). First-strand cDNA was synthesized using reagents provided in the SMART-PCR cDNA Synthesis Kit (Clontech). One tenth of first-strand cDNA was amplified with SMART-PCR amplification system (Clontech). Samples were cycled until the cDNA yield reached a plateau.

Probe synthesis

About 500 ng purified SMART double-strand (ds) cDNA and the Atlas Array gene-specific primer mix (1176 individual primers) were used to make the probe, using 33P deoxyadenosine triphosphate and Klenow enzyme according to the protocol supplied with the Atlas Arrays (Clontech). The probes were purified with the NucleoSpin Extraction Kit (Clontech).

cDNA arrays

Broad-coverage Atlas 1.2 Mouse Array (Clontech) was designed to profile the transcripts involved in most crucial cellular pathways and functions. The selection of genes was not tissue, function, pathway, or mechanism specific. The gene list and complete description of the array can be found at

Analysis of results

Membranes were scanned using a Phosphor-Imager (Molecular Dynamics, Sunnyvale, CA). The resulting images were then processed and analyzed using Atlas Image 1.5 Software (Clontech). AtlasImage normalization was carried out using the global method, which is preferable when closely related samples are compared (in fact, because of a homogeneous hybridization procedure, very similar results were obtained using the set of housekeeping genes on the membrane). The clustering was performed with AtlasNavigator software (Clontech), using the default standard correlation distance definition (0.95 correlation for the genes from the same cluster). Individual genes were examined and the clusters were refined using the “Find Similar” option of the software. Finally, primary data points were manually examined for all the candidate genes and those with the max value more than 3 times over the background were retained. The 2-tailed, paired Student t test (Microsoft Excel software; Microsoft, Seattle, WA) was used for statistical analysis.

RT-PCR analysis

RT-PCR was carried out using one tenth of the initial first-strand cDNA using advantage cDNA Polymerase Mix (Clontech) using Custom Atlas Array Primers. These primers are identical to the ones used to amplify the cDNA fragments immobilized on the arrays. Initially, 24 cycles of PCR amplification (30-μL reactions) were performed, and 4 μL was analyzed on a 2% Tris-acetate-EDTA (TAE) agarose gel. Additional cycles were performed if PCR product was not visible. The cDNA concentrations were normalized in all of the samples based on the RT-PCR of the ubiquitin gene, which was among the most abundantly expressed and whose expression well correlated with the global normalization values used for membrane analysis. The signals were digitized using the integrated Alpha Imaging Station Software (Alpha Innotech, San Leandro, CA).


HSCs and progenitor populations were prospectively isolated by flow cytometry using the appropriate marker set (Figure 1) as previously described.5,8,9,13 Because of the rare nature of the cells under investigation (ie, 0.05%-0.1% in the bone marrow), the purity of isolated populations was crucial for the analysis. To re-ensure the purity, all populations were isolated by flow cytometry (triple sorted), resulting in more than 99% purity as documented by the reanalysis after the third sort (Figure 2). About 4000 cells of each population were obtained to isolate total RNA and prepare the first-strand cDNA using SMART technology (Clontech). As few as 19 cycles of SMART-PCR amplification was sufficient to obtain the desired amount of amplified cDNA. This relatively low number of cycles ensures a minimal loss of transcripts and increases the likelihood of the linearity of the PCR-amplification step.

Despite the analysis of closely related populations, already at this point different expression patterns were clearly observed among the cDNA from HSCs and different progenitor populations, suggesting the presence of many strongly differentially expressed transcripts (Figure 3). For example, note the resemblance of the amplified cDNA patterns between all myeloid committed cells (ie, CMP, GMP, and MEP) the lymphoid committed cells (ie, CLP), and pro-B and pro-T cells on the other side. The pattern of amplified cDNA from HSCs clearly stands apart from all other patterns. PCR-amplified double-stranded cDNA probes were radio-labeled with 32P and hybridized to the array of 1176 known mouse genes, selected to represent most known cellular pathways and functions (1.2K Atlas Array; Clontech). The membranes were scanned and signals were analyzed.

Figure 3.

Gel electrophoresis of SMART-PCR amplification after 19 cycles on 1% agarose gel. The first lane is a 1-kb DNA ladder (Pharmacia, Piscataway, NJ). Note the standing-apart cDNA profile of HSCs and clear similarities between the lymphoid and myeloid committed progenitor populations.

Using standard distance correlation procedures, incorporated into the Atlas Navigator software, we identified and refined the gene clusters containing the transcripts that are enriched (at least 2 times over the average expression level) in each individual population analyzed (Figure 4). We have found the largest set of differentially expressed genes in the HSC cluster (43%), followed by the pro-T cluster (21%) (Figure 4A). Both CLPs and CMPs, the progenitor populations most closely related to HSCs, have the smallest numbers of differentially expressed genes (5% and 2%, respectively). We have observed a progressive increase in gene expression from HSCs to CMPs for most of the genes from the GMP cluster (Figure 4C) and to a lesser extent in the MEP cluster (Figure 4D). Similarly, we have seen a progressive up-regulation from HSCs to CLPs for most genes from the pro-T cluster (Figure 4F) and to a lesser degree in the pro-B cluster (Figure 4E). Thus, the genes associated with the most differentiated populations were already up-regulated in the corresponding progenitors undergoing cell fate transitions during hematopoietic differentiation (Figure 1). On the other hand, we have found some trends of a peculiar dichotomy of gene expression in various committed clusters. Namely, the pro-T—enriched genes were markedly decreased in pro-B cells, compared with CLP expression level (P = .0001, paired t test). Similarly, the GMP-enriched genes were decreased in MEP, compared with CMP expression levels (P = .003, paired t test). However, this trend was not observed for the genes enriched in pro-B cells and MEPs (P = .5, paired t test) for CLP versus pro-T and GMP versus CMP, respectively. We also have looked for the genes enriched in 2 or more populations. We have found one cluster of genes that is up-regulated in both CMP and GMP populations (Figure 4G). Few individual genes were up-regulated in 2 or more populations. For instance, both early B-cell factor (COE1) and lymphoid-enhancer binding factor 1 (LEF1) were found to be up-regulated in CLP and pro-B cells (Table 1).

Figure 4.

The results of gene clustering analysis across the compared populations. (A) Pie diagram of all differentially expressed genes; (B-G) bar graph representation of the HSC, GMP, MEP, pro-B, pro-T, and GMP + CMP clusters, respectively. The genes in each cluster are listed in Table 2. The inserts show average values for each population (± SD) in the same coordinates.

View this table:
Table 1.

Atlas cDNA array data for the reference genes

To scrutinize the hybridization data, we first examined several markers with known expression profiles (Table 1). The expression of c-kit mRNA in different populations closely correlated with the protein level of this marker used for isolation of the corresponding populations (Figure 1). We have found strong expression of interleukin-7 receptor α (IL-7Rα) in the CLP population; indeed, IL-7Rα is a marker for CLPs8 and was used for their prospective isolation (Figure 1). In agreement with the literature, a somewhat lower expression of IL7-Rα was found in pro-B cells.14 The low affinity Fcγ receptor, specifically distinguishing GMPs from CMPs and MEPs,9 was indeed found in the GMP cluster but not in the CMP and MEP clusters. Consistent with previous reports, EpoR and erythroid nuclear factor 2 (NF-E2) were highly expressed in MEPs, whereas GATA3 was greatly up-regulated in pro-T cells.9 As expected, erythroid Kruppel-like transcription factor (EKLF) was enriched in MEPs, and lymphocyte-specific tyrosine-kinase Lck was progressively up-regulated from HSCs to CLPs to pro-T populations. Accordingly, LEF1 was found enriched in the pro-B cluster, consistent with its role in pro-B but not in pro-T cells.15 High levels of β1 integrin mRNA found in HSCs in the present study are in agreement with the high levels of that protein found on purified HSCs compared with other bone marrow cells as detected by flow cytometry.16

To validate further the hybridization results, we performed a direct semiquantitative RT-PCR analysis for 6 genes selected from the HSC, CLP, CMP, and MEP clusters. A good correlation between the 2 methods was observed (Figure 5). Although somewhat lower contrast was seen with the RT-PCR analysis for IL-15 and insulin-like growth factor binding protein 4 (IGFBP4), this independent method (which does not include the SMART-PCR amplification step) confirmed the hybridization results. HSCs express high levels of c-mpl, a receptor for thrombopoietin (TPO), as detected by both methods. In fact, in our previous study we have found that the level of c-mpl in HSCs is much higher than in unfractionated whole bone marrow (WBM).7 It is somewhat surprising that c-mpl levels in CMP and MEP are relatively low, as both CMP and MEP respond to TPO to form megakaryocyte-erythroid colonies, and in that study c-mpl was highly expressed in MEP.9 Nevertheless, the high levels of c-mpl on HSCs support the physiologic role of c-mpl in regulation of HSC production and function. TPO and stem cell factor (SCF) were most efficient to induce in vitro cell division of HSCs,17-19 and TPO is 1 of only 3 cytokines shown to act alone to stimulate proliferation and differentiation of long-term HSCs (LT-HSCs) from H2K-BCL2 mice in serum-free medium.20 Indeed, the c-mpl-/- mice have deficiencies in multiple hematopoietic lineages and an intrinsic defect in HSC homing.21

Figure 5.

Comparison of the direct RT-PCR analysis versus hybridization with Atlas cDNA array.

A number of genes, some previously associated with hematopoietic stem cell self-renewal, were found to be differentially expressed in HSCs compared with various progenitor populations, suggesting their role in various HSC functions (Table 2). We and others previously reported that HSCs express high levels of angiopoietin 1.7,22 In this study we have found its receptor, an endothelial cell receptor tyrosine kinase Tie-2/Tek, also being enriched in HSCs. Tie2/Tek has been previously shown to play a critical role in definitive hematopoiesis.23-25 It will be important to test if HSCs at the single-cell level express both angiopoietin 1 and Tie2 implying an autocrine regulation, or if they are expressed by separable subsets.

View this table:
Table 2.

Annotated list of transcripts enriched in the compared populations corresponding to the clusters in Figure 4

We have found the multidrug resistance protein 1 (MDR1) efflux pump, previously proposed to be responsible for exporting rhodamine 123 and thus responsible for the rhodamine low phenotype, associated with the long-term repopulating activity in bone marrow.26,27 Moreover, the early growth response protein 1 (EGR1) transcription factor, which is necessary and sufficient to activate the expression from the MDR1 promoter element,28 was also found enriched in the HSC cluster. EGR1 has been shown to be induced by mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) pathway,29 and indeed, Map kinase kinase 3 (MEK3; MKK3) was found in the HSC cluster. Intriguingly, the key players of 2 other major signal transduction pathways, phospholipase C gamma and phosphoinositide-3(PI3)—kinase (catalytic subunit α) were also present in the HSC cluster. It is noteworthy that another component of PI3-kinase (the p55 regulatory subunit) was identified in the screen for the transcripts associated with the long-term repopulating HSCs (A.V.T. and I.L.W, unpublished observation, September 1998). We have found genes Eya1 and Eya2 in the HSC cluster. In addition to the vast array of developmental abnormalities,30 occasionally anemic embryos can be seen (P. Xu, personal oral communication, March 2000), suggesting a hematopoietic defect. We are currently investigating the hematopoietic deficiency in these mice.

Several general transcription factors were found in the HSC cluster. Namely, cAMP response element binding (CREB) binding protein (CBP), a coactivating transcription factor playing an essential role in hematopoiesis31; ATF4, a transcription factor belonging to the CREB/ATF family whose binding sites are present in a variety of growth-regulating cellular genes32; and a general transcription factor II-I, homolog of human BAP-135, also described as a target of Bruton tyrosine kinase.33 Several homeobox transcription factors HoxA7, HoxA9, and DM-locus associated homeodomain protein (DMAHP)/Six5 homolog were found in the HSC cluster. Indeed, the HoxA9 knock-out mice have defects in myeloid, erythroid, and lymphoid hematopoiesis,34 emphasizing the important role of HoxA9 in HSCs and/or function. Also, HoxA9-mediated immortalization of myeloid progenitors, together with cofactors Pbx and Meis point toward its potential role in self-renewal.35 We have recently suggested a role of Wnt/β-catenin pathway in HSC self-renewal.36 Indeed, we found a frizzled 6 homolog and axin enriched in HSCs, but not in committed progenitors. Found in the HSC cluster were 2 transcripts directly involved in the regulation of cell death. One, a recently identified receptor interacting protein (RIP) effector kinase, mediates an alternative caspase-8—independent cell death pathway triggered by Fas ligand.37 Another, a close homolog of BCL-2, the antiapoptotic protein myeloid cell leukemia 1 (MCL-1), was shown to be involved in leukemogenesis, regulation of apoptosis, and cell cycle progression.38-40 Enlarging our previous finding of Spi2A serpin protease inhibitor being highly enriched in HSCs,7 Spi3 and neuroserpin were found enriched in HSCs in the present study.

In contrast, most genes enriched in committed progenitors were associated with various aspects of lymphocyte or monocyte differentiation, immune response, and metabolism (Table 2). Several genes such as IL-2 receptor gamma chain, Fas ligand, Lck, ZAP-70, c-Fos, c-Jun, A20, and granzyme A, typically associated with T-cell differentiation, were indeed found in the pro-T cluster. For most of them a progressive up-regulation was seen from HSCs to CLPs to the pro-T cluster and down-regulation in the pro-B cluster. The finding of macrophage CSF1 (M-CSF1) receptor c-fms proto-oncogene enriched in CMPs is consistent with the fact that it was often found in monocytes/macrophages, seldom in erythroid and megakaryocytic lineage cells, and never in lymphocytes.41 The presence of the histidine decarboxylase (HDC) gene in the GMP cluster is consistent with the fact that the macrophage lineage in mouse bone marrow stains positive for histidine decarboxylase (HDS) protein and that HDC is involved in human monocyte/macrophage differentiation.42,43 Patients with Friedreich ataxia have relative heme deficiency leading to a state of ineffective persistent erythropoiesis,44 and the frataxin gene was found enriched in the MEP cluster.


Here we present a systematic comparative study of gene expression in pluripotent HSCs capable of long-term self-renewal and various multipotent short-lived committed lymphoid and myeloid progenitors. There are at least 2 major differences between HSCs and committed progenitors: multipotency versus oligopotency and life-long self-renewal versus the absence of self-renewal. Various genes involved in lineage commitment would be expected to be progressively up-regulated from CLPs to pro-B or pro-T cells and from CMP to GMPs or MEPs. Indeed, we have found many examples of those; in fact most of the genes in pro-T, MEP, and to a lesser extent in pro-B and GMP clusters follow this rule (Figure 4C-F). On the other hand, genes involved in extensive self-renewal would be expected to be associated only with HSCs and not progenitor populations.

We have found major quantitative and qualitative differences in gene expression between HSCs and all other progenitors by clustering the genes selectively associated with each population. Namely, we have found that almost half of all differentially expressed genes in this analysis were associated with HSCs; in contrast very few differentially expressed genes were found in the close intermediate progenitor populations CLP and CMP. In the cluster representing HSC-associated genes, most were developmentally regulated; some were involved in general transcription activation, cell cycle, cell death, or signal transduction. On the contrary, most of the genes associated with various progenitor populations are implicated in lymphoid or myeloid differentiation, immune response, or metabolism.

In particular, we have found several transcriptional integrators and coactivators (CBP, ATF4, general transcription factor II-I) as well as the key elements of 3 major signal transduction pathways (phospholipase C—gamma [PLC-gamma], PI-3 kinase, and MEK3/MKK3) are enriched in the HSC pool. Increased levels of the above-mentioned genes and the fact that HSCs have far more differentially expressed genes than any closely related progenitor population argue in favor of a high level of regulated transcriptional activity in HSCs. That supports a view of stem cells as being rather “active” cells, capable of sampling a large number of environmental cues and translating them into signals allowing decisions that maintain the balance between the self-renewal and differentiation.

Even with the limited set of genes analyzed, these findings implicate several of the putative pathways that operate in HSCs. Namely, both the EGR1 transcription factor and its target gene, the MDR1 efflux pump (involved in the rhodamine 123 efflux from stem cells), were found associated with HSCs. Moreover, MEK3, a major player in the ERK/MAPK pathway known to activate EGR1,29 was found in the HSC cluster. It is noteworthy that another efflux pump, the ABC transporter breast cancer resistance protein 1 (Bcrp1), was recently found to be highly expressed in several populations enriched for stem cells and is believed to be the molecular determinant of the Hoechst 33342 dye efflux characteristic of the side population.7,45

We have previously identified the transcripts enriched in HSCs compared with the adult whole bone marrow.7 Many of the HSC-enriched genes identified in that early study were not present on the Atlas 1.2 Mouse Array (eg, Cyt28, mETL1, Spi2a). Nevertheless, in the present study, analyzing much narrower intervals of hematopoietic differentiation, we have found a number of genes in the HSC cluster that are known to interact with gene products identified in our previous study. For instance, we have previously found angiopoietin 1 transcript expressed in HSCs at a high level compared with the rest of the bone marrow.7 In the present study we have shown that its receptor Tie-2/Tek is selectively enriched in HSCs. Indeed, it was suggested that angiopoietin 1 and vascular endothelial growth factor (VEGF) stimulate postnatal hematopoiesis by recruitment of vasculogenic and hematopoietic stem cells.46 Likewise, we have previously found high levels of Meis1 transcript in HSCs.7 In the present study we found 3 homeobox genes Hox-A7, HoxA9, and DMAHP/Six5 associated with HSCs. Indeed, it was previously shown that HoxA9-mediated immortalization of myeloid progenitors requires functional interactions with cofactors Pbx and Meis.35 Thus, our finding of HoaA9 in the HSC cluster is consistent with the previous results and points toward the possible role of Meis1/Pbx/HoxA9 (HocA7, DMAHP/Six5) complex in self-renewal of HSCs.

The higher expression of c-mpl in HSCs than in cells contributing to megakaryocyte cell fate (CMP, GMP) deserves special mention. Although the ligand for c-mpl is called thrombopoietin, implying a platelet-specific cytopoietic effect, in fact TPO is an important early-acting cytokine to induce human HSC proliferation.17 Furthermore, single highly purified LT-HSCs that have enforced expression of hBCL-2 can respond to only 3 known single cytokines in serum-free medium by proliferation and differentiation—SCF, TPO, and IL-3.20 Each has a different outcome; SCF leads to all progenitors, IL-3 to myeloid only, and TPO to megakaryocytic.20 This is consistent with a role for TPO in helping specify outcomes at the HSC level. It does not act only at the HSC level, as CMP and MEP require TPO to give rise to megakaryocytes.9 Thus, our array analysis of purified stem and progenitor cells can reveal unexpected pathways in hematolymphoid differentiation.

Our data suggest that HSCs are highly transcriptionally active cells possibly with a significant amount of open chromatin. The differentiation of HSCs toward a particular lineage is characterized by shrinking of the repertoire of expressed genes, down-regulation of HSC-enriched genes, and up-regulation of a relatively small but defined subset of lineage-specific genes, associated with the function of terminally differentiated cells.


We thank Koichi Akashi for his help in cell isolation, Drs Peter Simonenko and Michael Makhanov for help with data analysis, and Libu Jerabek for laboratory management and technical help. A.V.T. was an Irvington Institute fellow.


  • Reprints:
    Alexey V. Terskikh, Department of Life Science, Swiss Federal Institute of Technology Lausanne (EPFL), CH-1015 Lausanne, Switzerland; e-mail: alexey.terskikh{at} and terskikh{at}
  • Prepublished online as Blood First Edition Paper, March 13, 2003; DOI 10.1182/blood-2002-08-2509.

  • Supported by National Institutes of Health grant PO1DK53074-06 (I.L.W.).

  • 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.

  • Submitted August 23, 2002.
  • Accepted January 28, 2003.


View Abstract