The Kinetic Status of Hematopoietic Stem Cell Subpopulations Underlies a Differential Expression of Genes Involved in Self-Renewal, Commitment, and Engraftment

Authors


Abstract

The gene expression profile of CD34 hematopoietic stem cells (HSCs) and the correlations with their biological properties are still poorly understood. To address this issue, we used the DNA microarray technology to compare the expression profiles of different peripheral blood hemopoietic stem/progenitor cell subsets, lineage-negative (Lin) CD34, LinCD34+, and Lin+CD34+ cells. The analysis of gene categories differentially expressed shows that the expression of CD34 is associated with cell cycle entry and metabolic activation, such as DNA, RNA, and protein synthesis. Moreover, the significant upregulation in CD34 cells of pathways inhibiting HSC proliferation induces a strong differential expression of cyclins, cyclin-dependent kinases (CDKs), CDK inhibitors, and growth-arrest genes. According to the expression of their receptors and transducers, interleukin (IL)-10 and IL-17 showed an inhibitory effect on the clonogenic activity of CD34 cells. Conversely, CD34+ cells were sensitive to the mitogenic stimulus of thrombopoietin. Furthermore, CD34 cells express preferentially genes related to neural, epithelial, and muscle differentiation. The analysis of transcription factor expression shows that the CD34 induction results in the upregulation of genes related to self-renewal and lineage commitment. The preferential expression in CD34+ cells of genes supporting the HSC mobilization and homing to the bone marrow, such as chemokine receptors and integrins, gives the molecular basis for the higher engraftment capacity of CD34+ cells. Thus, the different kinetic status of CD34 and CD34+ cells, detailed by molecular and functional analysis, significantly influences their biological behavior.

Introduction

A novel class of hematopoietic stem cells (HSCs) lacking the CD34 protein has been described in mice and humans [1]. In vitro and in vivo studies have led to the hypothesis that HSCs exist in two functional states that can be distinguished by CD34 expression.CD34cellsrepresentareservoirofkineticallyandfunction-ally resting HSCs that need to be activated to generate a CD34+ cell population with high proliferative and engraftment potential [2, 3]. LinCD34 cells seem to be mainly out of cycle and have minimal, if any, colony-forming and long-term culture-initiating cell (LTC-IC) ability. Conversely, most mobilized CD34+ cells are in G1 phase and retain clonogenic and LTC-IC activity [3]. In vivo, LinCD34 cells derive from CD34+ progenitors and regain expression of CD34 on secondary transplantation [4].

The microarray technology has been recently used to find the correlations existing between the gene expression profile and the human HSC biology by the comparison of CD34+ HSCs obtained from different sources [5] and subsets [6, 7]. Furthermore, the genome-wide analysis provides a valuable tool for examining how the genetic programs underlying the self-renewal and commitment [8, 9] are established in normal hematopoiesis.

Although some studies in mouse and in human have suggested that the CD34 expression correlates with cell proliferation [3, 10, 11], the relationship between the expression of CD34, cycling status, self-renewal, and lineage commitment is still poorly understood.

In this study, we attempted to address these issues by evaluating the gene expression profile of different subsets of peripheral blood hemopoietic stem/progenitor, LinCD34, LinCD34+, and Lin+CD34+ cells. Our data indicate that the CD34/CD34+ transition is associated with cell cycle recruitment, metabolic activation, and downregulation of growth-inhibitory pathways. The differential activation of these pathways leads to a strong differential expression of cyclins, CDKs, CDK inhibitors, and growth-arrest genes between CD34 and CD34+ cells. Moreover, CD34+ cells show a significant upregulation of self-renewal, commitment, and engraftment-related genes [3, 12], whereas LinCD34 cells express preferentially genes related to quiescence and to neural, epithelial, and muscle differentiation.

Materials and Methods

Cells

To obtain peripheral blood hemopoietic stem and progenitor cells, 48 healthy donors received recombinant human (rh) G-CSF (Lenograstim, Rhone-Poulenc Rorer, Milan, Italy), administered subcutaneously at 10 μg/kg per day for 5–6 days. Hemopoietic stem/progenitor cell purification and phenotypic analysis were performed as previously described [3, 13]. Lin+CD34+ cells were purifiedfrom24donors, whereas LinCD34+andLinCD34cells were simultaneously purified from 24 additional donors. Aliquots of purified Lin+CD34+, LinCD34+, and LinCD34 cells were reanalyzed by FacScan (Becton, Dickinson, Franklin Lakes, NJ) to assess their purities, which were 98.2% ± 0.5%, 99.1% ± 0.7%, and 99.8% ± 0.1%, respectively. A representative example of sorting gates and flow cytometry reanalysis of sorted cells is shown in supplementary online Figure 1. In addition, FACS analysis of CD45 antigen (Ag) on highly purified LinCD34 HSCs confirmed their hematopoietic origin (supplementary online Fig. 2).

Progenitor Cell Assays

Human colony-forming unit (CFU) cells were cultured in methylcellulose, as previously reported [3], with and without 200 ng/ml of rh interleukin (IL)-17 or 100 ng/ml of rh IL-10 (Biodesign, Saco, ME). CFU-GM, BFU-E, and multilineage colonies (CFU-Mix) (together referred to as CFU-C) were scored after 14 days. Megakaryocyte progenitors (CFU-MK and BFU-MK) were assayed and identified in plasma-clot culture as previously described [14]. In brief, CFU-MK cells were identified after 12 days of culture by fixing plasma clot in situ with methanol-acetone 1:3 for 20 minutes; washing with phosphate-buffered saline and double distilled water; and then air drying. BFU-MK cells were fixed after 19 to 21 days of culture. After immunofluorescent staining with the monoclonal antibody CD41a (Dako, Glostrup, Denmark) directed against the glycoproteic αIIb-β3 complex, CFU-MK cells were scored as aggregates of three or more intensely fluorescent cells, and BFU-MK consisted of two to six fluorescent foci containing more than 100 cells.

Clonogenic assays were performed on freshly isolated LinCD34+ or LinCD34 cells plated at the concentration of 104 cells per ml or seeding the same cell populations after 7, 14, and 21 days of incubation in liquid culture on a feeder layer (see below). Additionally, the clonogenic activity of megakaryocyte progenitors was assessed after incubation of LinCD34+ and LinCD34 cells in serum-free medium added with TPO (100 ng/ml; Amgen, Thousand Oaks, CA) for 19–21 days (see below).

The results of clonogenic assays are expressed as the mean ± standard deviation of at least three different experiments. Results of in vitro studies were analyzed with the paired nonparametric Wilcoxon rank-sum test. Two-sided p values lower than .05 were considered statistically significant.

Liquid Cultures

Forty thousand hematopoietic LinCD34+ and LinCD34 cells per ml were cultured onto irradiated murine stromal cells (M2-10B4) genetically engineered to produce G-CSF and IL-3 as described elsewhere [13]. Cultures were supplemented with optimized concentration of the following rh cytokines: stem cell factor (50 ng/ml), IL-11 (50 U/ml; Endogen, Woburn, MA), FLT-3L (50 ng/ml; Immunex, Seattle), and TPO (100 ng/ml) with and without rh IL-17 or rh IL-10 (200 and 100 ng/ml, respectively). Ten thousands cells were cultured at weekly intervals in methylcellulose to evaluate the presence of secondary CFU-C. Furthermore, to assess the percentage of CD34+ cells in liquid culture, the cells were incubated with anti-human CD34-fluorescein isothiocyanate monoclonal antibody (HPCA-2; Becton, Dickinson, San Jose, CA) and then analyzed by a FACScan equipment [3].

To determine the activity of TPO in stimulating hematopoietic progenitor cells in liquid culture, freshly isolated LinCD34+ and LinCD34 cells from three healthy subjects were resuspended in serum-free medium (X-Vivo 20, Bio-Whittaker, Walkersville, MD) in the presence of 100 ng/ml of TPO alone and cultured for 19–21 days. Cell cultures were added with fresh serum-free medium and 100 ng/ml of TPO every 3 days. At the end of the culture, 8 to 10 × 104 cells per ml were then assayed in plasma clot to evaluate the presence of CFU-MK and BFU-MK. The percentage of CD34+ cells in liquid cultures was also assessed as reported above.

RNA Extraction and Microarray Data Analysis

Total RNA was isolated from each cell population (2 × 105 cells) of each donor using a modification of the guanidinium isothiocyanate procedure and ultracentrifugation on cesium chloride gradient [15]. Disposable RNA chips (Agilent RNA 6000 Nano LabChip kit, Agilent Technologies, Waldbrunn, Germany) were used to determine the concentration and purity/integrity of RNA samples using Agilent 2100 bioanalyzer.

RNAs originating from 12 donors were pooled to obtain at least 2 μg per sample. A replicate experiment was carried out in a similar way on 12 different normal donors. The biotin-labeled target synthesis reactions, as well as the Affymetrix HG-U95Av2 GeneChip arrays hybridization, staining, and scanning, were performed, starting from 2 μg of total cellular RNA, using the Affymetrix standard protocols (Affymetrix, Santa Clara, CA).

The MAS 5.0 absolute analysis algorithm was used to determine the amount of a transcript mRNA (signal), whereas the MAS 5.0 comparison analysis algorithm was used to compare gene expression levels between two samples.

Differentially expressed genes were selected as the sequences showing a change call I or D at least once in the pairwise comparisons between each replicate and the other cell populations. The genes passing this filter were selected as the changing genes. Genes showing a detection call A (absent) in all samples were excluded [16]. The generated list and, independently, the MAS 5.0–generated absolute analysis data were uploaded onto Gene-Spring software version 6.1 (Silicon Genetics, Redwood City, CA). To normalize data, each measurement was divided for the 50th percentile of all signals in that sample. The percentile was calculated with all normalized signals above 10. Each gene was divided by the median of its measurements in all samples.

Then, using the GeneSpring package filtering options, poorly changed genes (i.e., those showing a normalized intensity between 0.7 and 1.33) were filtered out from the changing genes list described above. Among these sequences, a Welch analysis of variance test (parametric test, with variances not assumed equal, p-value cutoff .05, multiple testing correction: Benjamini and Hochberg false discovery rate) passed 2,720 changing and reliable sequences. This list underwent clustering analysis using the analysis options (gene trees and condition trees) included in the GeneSpring package, applying different correlation equations.

To improve our ability to interpret the biological meaning of microarray data, we used GenMAPP 1.0 software (Gene Micro-Array Pathway Profiler; www.genmapp.org). To identify the gene ontology (GO) categories characterized by significant number of genes differentially expressed in each cell population, we used an accessory program of GenMAPP, MAPPFinder 1.0 beta [17].

Real-Time Quantitative Polymerase Chain Reaction

cDNA was reverse transcribed from total RNA samples (100 ng per sample) obtained from six additional healthy donors using High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA) as described in the manufacturer's protocol. TaqMan polymerase chain reaction (PCR) reactions were carried out from cDNA samples using the TaqMan Universal PCR Master Mix (Applied Biosystems), as described in the manufacturer's protocol, onto custom 7,900 micro-fluidic cards (Applied Biosystems) by means of ABI PRISM 7900 HT Sequence Detection Systems. TaqMan strategies for each gene were developed as Assay on Demand by Applied Biosystems. Gene expression profiling was achieved using the comparative cycle threshold (CT) method of relative quantification. To normalize data, for each sample, delta-delta-CT was calculated using the median of its delta-CTs in all samples as calibrator. Normalized delta-delta CTs were than uploaded onto GeneSpring using the real-time data transformation.

Results

Global Transcriptome Changes in LinCD34, LinCD34+, and Lin+CD34+

We assessed, in duplicate, the gene expression in all three hemopoietic stem/progenitor cell populations using Affymetrix HG-U95Av2 GeneChip array, representative of 12,625 transcripts.

All of the data have been deposited in the Gene Expression Omnibus MIAME-compliant public database at http://www.ncbi.nlm.nih.gov/geo. LinCD34 accession numbers are GSM25887 and GSM25888 (I and II replicate, respectively), LinCD34+ accession numbers are GSM25885 and GSM25886 (I and II replicate, respectively), and Lin+CD34+ accession numbers are GSM25883 and GSM25884 (I and II replicate, respectively).

The mRNA complexity significantly increased upon the acquisition of CD34 Ag expression; in fact, 5,348 versus 4,524 sequences are called present by Affymetrix MAS 5.0 absolute analysis algorithm in LinCD34+ and LinCD34 HSC, respectively; mRNA complexity increased also during LinCD34+ and Lin+CD34+ transition (6,128 versus 5,348 sequences called present in Lin+CD34+ and LinCD34+ HSCs, respectively). The most significant transcriptome changes were found between LinCD34 and Lin+CD34+ cells (supplementary online Table 1).

Clustering Analysis of Genes Differentially Expressed in Hematopoietic Stem/Progenitor Cells

A list of 2,720 changing and reliable genes that obtained filtering data as described in Materials and Methods was used for hierarchical clustering. The condition tree shows that the clustering algorithm hierarchically paired the two CD34+ population transcript profiles (Fig. 1).

Figure Figure 1..

Clustering of the 2,720 most changing genes. Clustering has been performed using an unsupervised approach and applying several clustering algorithms provided by GeneSpring. A combination of two hierarchical clustering analyses (gene tree and condition tree) is shown. The gene tree is shown on left; the condition tree is shown on top. Gene coloring was based on normalized signals as shown at the bottom of the figure.

GO Mapping of Differentially Expressed Genes

To identify whether the differentially expressed genes underlie a prevalent biological process, we uploaded the gene list of 2,720 changing and reliable genes onto MAPP Finder software. The prevalent categories in the biological process GO tree include protein biosynthesis, cell cycle, RNA metabolism, DNA replication and chromosome cycle, chromatin assembly/disassembly, tricarboxylic acid (TCA) cycle, DNA repair, oxidative phosphorylation, ubiquitin-dependent protein degradation, and transcription from Pol II promoter (supplementary online Table 2).

Other categories, not evidenced by GO mapping analysis, had been examined, such as cell adhesion, cytokine and hematopoietic growth factor receptors, and transcription factors.

Cell Cycle Regulator Gene Expression

The analysis of expression of cyclins, CDKs, cyclin-dependent kinase inhibitors (CDKNs), and growth-arrest genes led to the following results (Fig. 2A). First, CD34 cells were characterized by a preferential expression of growth arrest genes, such as Gas6, RGS2, ZFP36, ING1, PEDF, and LNK [18, 19] and of some CDKNs, such as CDKN1A (p21 waf-1) [20], CDKN2C (p18) and CDKN2D (p19) [21]. This expression pattern was paralleled by a concomitant lower expression of cyclins and CDKs compared with CD34+ cells. Second, CD34+ cells preferentially expressed early G1 cyclins and CDKs (D2 and D3 cyclins, CDKs 4 and 6). Very low levels of late G1 or mitotic cyclins and associated CDKs were also detected. Moreover, some cell cycle–related genes, such as NFY and c-myb, which regulate the promoter activity of the human CD34 gene [10], were shown to be increased in the CD34/ CD34+ transition (see below).

Figure Figure 2..

Expression of cell cycle regulators, growth factors receptors, and cytokines. Eisen tree map computed using the GeneSpring gene tree and the Pearson correlation equation on the modulated probe sets belonging to the following categories: (A) cell cycle, (B) growth factor receptors, and (C) cytokines. The signal-based coloring legend is shown at the bottom of the figure.

Differential Activation of Pathways Leading to Cell Proliferation or Growth Inhibition

To better characterize the molecular mechanisms regulating HSC proliferation or quiescence, we assessed whether growth factor receptors were differentially expressed in CD34+ and CD34 cells (Fig. 2B). The results obtained evidenced that distinct sets of receptors are preferentially expressed by CD34 (IL-17R, transforming growth factor [TGF]-βR1, TGFβR2, IL-10RA, IL-10RB) or CD34+ cells (FLT3, MPL, IFNγR2, EpoR).

We then analyzed the expression of genes involved in TPO, FLT3, IFNγ, IL-17, IL-10, and TGFβ pathways by graphic visualization of gene expression using GeneMAPP software.

Our data showed that signal transducers, particularly those involved in Ras-mediated pathways, are constitutively expressed in all populations studied, whereas receptors, primary response genes, and inhibitors are differentially expressed as follows.

TPO Pathway

This pathway is apparently upregulated in CD34+ cells, because TPO response genes, such as HOXB4, CBFß, Runx1, Nfe-2, Gata-2, Fli-1, CD41b, CD42b, and CD61, are upregulated in this cell population (Fig.3A). Consistently, CD34+cells gave origin to early and late megakaryocyte progenitors (i.e., CFU-MK and BFU-MK) in response to TPO, whereas CD34 cells did not show any colony-forming ability even after 7 days of culture (Figs. 4A, 4B). Interestingly, cultures of CD34 cells generated secondary CFU-MK after 21 days in serum-free medium in the presence of TPO.

Figure Figure 3..

Visualization of expression data on the maps of (A) TPO, (B) IL-10, and (C) IL-17 pathways. Genes are colored according to the absolute and comparative expression (Lin+CD34+ versus LinCD34 cells). The legend of the coloring criteria is reported on the right of the figure. Abbreviations: IL, interleukin; TPO, thrombopoietin.

Figure Figure 4..

Effects of TPO, IL-10, and IL-17 on LinCD34 and LinCD34+ cells. (A): TPO treatment. (a): TPO induces the clonogenic growth of CFU-MK and BFU-MK from freshly isolated CD34+ but not LinCD34 cells. (b): Similarly, 7 days of culture of LinCD34 cells onto irradiated murine stromal cells (M2-10B4), genetically engineered to produce G-CSF and IL-3 did not induce secondary colony formation in response to TPO. The results shown derive from six different experiments and are expressed as mean ± standard deviation. (c): Megakaryocyte progenitors (CFU-MK) became detectable after 19–21 days (day +21) of incubation of LinCD34 cells in serum-free liquid medium added with TPO. At this time point, 7% ± 3% of total cell population was represented by CD34+ cells (see text). (B): IL-10 treatment. Clonogenic efficiency of highly purified (a) LinCD34+ and (c) LinCD34 cells. The cells were cultured in semisolid medium for 14 days with GM-CSF, IL-3, EPO, and SCF. (b): LinCD34+ and (d) LinCD34 cells were cultured onto M2-10B4 stromal cells in the presence of a cytokine cocktail (FLT-3L, TPO, IL-11, and SCF) with or without IL-10 or in the presence of IL-10 alone. Subsequently, clonogenic assays in semisolid medium with GM-CSF, IL-3, EPO, and SCF were set up to assess the production of secondary CFU-C. The results are expressed as mean ± standard deviation of three different experiments. IL-10 showed no significant activity on clonogenic CD34+ cells (p = not significant), whereas the addition of IL-10 inhibited the secondary colony-forming activity of CD34 cells induced by cytokines in liquid cultures (p < .03). (C): IL-17 treatment. Clonogenic efficiency of highly purified (a) LinCD34+ and (c) LinCD34cells. The cells were cultured in semisolid medium for 14 days with GM-CSF, IL-3, EPO, and SCF. (b): LinCD34+ and (d) LinCD34 cells were cultured onto M2-10B4 stromal cells in the presence of a cytokine cocktail (FLT-3L, TPO, IL-11, and SCF) with or without IL-17 or in the presence of IL-17 alone. Subsequently, clonogenic assays in semisolid medium with GM-CSF, IL-3, EPO, and SCF were set up to assess the production of secondary CFU-C. The results are expressed as mean ± standard deviation of three different experiments. Similarly to IL-10, IL-17 showed no activity on clonogenic CD34+ cells (p = not significant), whereas the study cytokine inhibited the secondary clonogenic activity of CD34 cells induced by cytokines in liquid cultures (p < .04). Abbreviations: CFU-MK, colony-forming unit megakaryocyte; EPO, erythropoietin; IL, interleukin; SCF, stem cell factor; TPO, thrombopoietin.

At that time point, phenotypic analysis demonstrated the presence of CD34+ cells (7% ± 3% of the total population) deriving from CD34 HSCs (Fig. 4A, panel c). Although we have no formal evidence, it is conceivable that functional response to TPO is associated with the acquisition of the CD34 Ag.

IL-10 Pathway

IL-10 inhibits cell-cycle progression of HSCs and progenitor cells acting through STAT1/STAT3 activation [22]. Analysis of the genes involved in this pathway showed that both receptor isoforms (IL-10RA and IL-10RB) and IL-10 primary response genes, such as CDKN1A/p21 and CDKN2D/p19, were preferentially expressed in CD34 cells (Fig. 3B). Clonogenic assays on Lin cells demonstrated a minimal activity of IL-10 on CD34+ cells (Figs. 4B, panel b, 4B, panel b), whereas the addition of IL-10 inhibited the secondary colony-forming activity of CD34 cells induced by cytokines in liquid cultures (Fig. 4B, panel d) (p < .03). At this stage, the percentage of CD34+ cells deriving from CD34 HSCs after 7 days in liquid culture was 6% ± 2%, with no significant difference between IL-10–treated and control samples.

IL-17 Pathway

As reported in Figure 2, IL-17 receptor is specifically expressed by CD34 cells. Moreover, some IL-17 primary response genes (i.e., ICAM-1 and IL-8) are preferentially expressed in the same cell population (Fig. 3C). Consistent with these findings, we did not observe any significant activity of IL-17 on CD34+ cells (Figs. 4C, panel a, 4C, panel b) (p = not significant). In addition, we demonstrated for the first time the inhibitory effect of IL-17 on the secondary clonogenic activity of CD34 cells induced by cytokines in liquid cultures (Fig. 4C, panel d) (p < .04).

In this set of experiments, the percentage of CD34+ cells originating from CD34 cells after 7 days was 4% ± 2%, with no difference between IL-17–treated and control samples.

Finally, gene expression data suggest that FLT3 and IFNγ [23] pathways are active mainly in CD34+ cells, whereas TGFβ exerts its inhibitory effect preferentially on CD34 cells [24] (supplementary online Fig. 3).

Metabolic Activation of CD34+ Cells

Genes involved in DNA replication, such as DNA polymerases, topoisomerases, and minichromosome maintenance (MCM), were preferentially expressed in CD34+ cells, particularly in Lin+CD34+ cells (supplementary online Fig. 4). These data are consistent with the kinetic status of the three analyzed cell populations; in fact, CD34+ cells are mainly in G1 phase of cell cycle and synthesize the enzyme components for the subsequent S phase.

Genes involved in DNA repair (base excision repair, nucleotide excision repair, mismatch repair, and double-strand break repair) exhibited a preferential expression in CD34+ cells, particularly in Lin+CD34+ (supplementary online Fig. 5). Again these results are consistent with the kinetic and differentiation status of CD34 and CD34+ cells [25]. The global expression analysis of genes involved in RNA splicing, capping, and polyadenilation showed that the process of RNA maturation is mainly active in CD34+ cells (supplementary online Fig. 6). Moreover, combined analysis of the expression of genes codifying for ribosomal proteins demonstrated that these transcripts, already present in CD34 cells, undergo a remarkable induction in LinCD34+ and a slight decrease in the subsequent transition to Lin+CD34+ cells (supplementary online Fig. 7A). These data are in keeping with previous studies describing the increase of ribosome biogenesis during the G0/G1 transition [26]. The expression of genes involved in protein translation and modification was increased in CD34+ cells, particularly in Lin+CD34+ cells (supplementary online Figs. 7B, 7C).

Genes codifying for proteins involved in oxidative phosphorylation and TCA cycle processes showed a prevalent expression in CD34+ cells, especially in Lin+CD34+ (supplementary online Figs. 8A, 8B). These data are consistent with the already described activation of oxidative phosphorylation and TCA cycle in early G1 phase of the cell cycle [27].

Self-Renewal Capacity

Analysis of TF expression indicated that most genes involved in self-renewal process were upregulated in CD34+ cells (Fig. 5A). Among them, HOXA5, HOXA9, HOXA10, HOXB2, HOXB5, Meis1, and PBX2 are preferentially expressed by CD34+ cells.

Figure Figure 5..

Expression of transcription regulators and differentiation markers. Eisen tree map computed using the GeneSpring gene tree and the Pearson correlation equation on the modulated probe sets belonging to the following categories: (A) transcription factors, (B) transcription activators and repressors, and (C) differentiation markers. The signal-based coloring legend is shown at the bottom of the figure.

The expression of HOXB4, recently described as a key regulator of TPO-induced HSC self-renewal [28], was sixfold upregulated in LinCD34+ and ninefold upregulated in Lin+CD34+ compared with LinCD34 cells. HOXB4 expression was detected only by real-time quantitative PCR (RTQPCR), because the HOXB4 probe set is not represented on HG-U95Av2 array. GATA-2 and Bmi1, key factors for HSC self-renewal [29, 30], and LMO2, CBFß, and CUTL1, known regulators of early hematopoiesis [31], were found to be expressed in all cell populations, particularly in CD34+ cells. Conversely, ID1 and ID2 (inhibitors of cell differentiation) [32] resulted in downregulation of CD34+ cells. Genes belonging to the WNT and NOTCH pathways were expressed at very low levels in all cell populations.

Lineage Commitment Capacity

The expression analysis of TFs involved in all hematopoietic lineage differentiation (Fig. 5A) evidenced that several genes increased in CD34/CD34+ transition: GATA-1, PU.1/SPI1 [31], and HOXA5 [33] (myeloid commitment); GATA-1, LMO2, TAL-1, LDB1, and TCF3 [34] (erythroid commitment); GATA-1, CBFß,GATA-2, FLI1, and NF-E2 [35] (megakaryocytic commitment); GATA-1 and C/EBPß [36] (eosinophil commitment), and PU.1 and C/EBPß [37] (neutrophil commitment).

Although transcripts of TFs involved in monocyte (ICSBP1 [38], EGR-1 [39], HOXA10 [40]) and lymphoid (Ikaros, GATA-3 [41]) commitment were always detectable, variations of their expression levels did not correlate with the differentiation degree of the analyzed cell populations (Fig. 5A).

The upregulation of lineage-commitment TFs in CD34+ cells was associated with the induction of a large number of intracellular and surface markers belonging to the monocytic (CD14, ACO2, TMSF7), granulocytic (CD16, LILRA3, LILRB3, MMP9, CSF3R, CD32), lymphoid (CD69, CD19, CD164, CD58, TRB), megakaryocytic (PF4, F2R, VEGF, CD31, CD41, CD151, GP1BB), and erythroid (KLF1, RUVLB1, RUVLB2, GYPC) differentiation lineages (Fig. 5C).

Transcriptional activators, such as topoisomerase, helicases, acetyltransferase, and chromatin remodeling proteins (SMARCA2, SMARCD2, SMARCA4, SMARCC2, and SMARCC1) were mainly expressed in CD34+ cells; conversely, transcriptional repressors were preferentially expressed in LinCD34 (Fig. 5B).

Engraftment Capacity

Previous reports demonstrated that human CD34+ cells have a significantly greater engraftment potential than CD34 cells when transplanted in irradiated nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice [3, 12]. Our molecular analysis showed that the expression of genes belonging to the cell adhesion category is higher in CD34+ cells (supplementary online Fig. 9). Furthermore, genes specifically involved in the homing and engraftment of HSCs in the BM were preferentially expressed in CD34+ cells. In fact, LinCD34+ cells showed a higher expression of VLA-4, VLA-5, and SELPLG compared with LinCD34 cells (Fig. 6). Interestingly, CD34 cells exhibited higher levels of CXCR4, but also of RGS1 and RGS13, that function as negative regulators of CXCR4 activity by the inhibition of trimeric G proteins [42] (Fig. 6). Taken together, these observations support the view that CD34+ cells have higher engraftment capacity in primary recipients of xenogenic transplant compared with CD34 cells.

Figure Figure 6..

Schematic view of the engraftment pathway. Genes are colored according to the absolute and comparative expression (LinCD34+ versus LinCD34 cells). The legend of the coloring criteria is reported at the bottom of the figure.

Differential Expression of Nonhematopoietic Markers

Global expression analysis of nonhematopoietc markers, such as epithelial, neural, and muscle tissue markers, revealed their preferential expression in CD34 cells (Fig. 7). The expression of these genes strongly decreases during the CD34/CD34+ transition and becomes undetectable in terminally differentiated cells (R. Manfredini et al., unpublished data). Among these genes, the more differentially expressed were the epithelial markers CDH1 (E-cadherin) and K5 type II keratin (KRT5), the neural marker dopamine receptor 4 (DRD4), and the muscle marker tropomyosin 2, beta (TMP2).

Figure Figure 7..

Expression of nonhemopoietic markers. Eisen tree map computed using the GeneSpring gene tree and the Pearson correlation equation on the modulated probe sets belonging to the following categories: (A) neural markers, (B) muscle markers, and (C) epithelial markers. The signal-based coloring legend is shown at the bottom of the figure.

Real-Time Quantitative PCR Validation of Differential Expressed Genes

To confirm microarray data, we carried out a TaqMan RTQPCR analysis on a validation set containing 77 transcripts selected among the differentially expressed genes with greatest biological significance. TaqMan data were uploaded onto GeneSpring software as described in Materials and Methods and analyzed together with the array data. Supplementary online Figure 10 shows a gene tree and condition tree computed onto the validation set gene list using the Spearman correlation. Almost all of the analyzed genes showed the same expression pattern with both analyses.

Discussion

Although the molecular basis of human HSC functions, such as self-renewal, commitment, engraftment, and the so-called plasticity capacities, have been recently approached by DNA microarray technology [57], the molecular mechanisms underlying the biological properties of different HSC subsets, such as the LinCD34 cell population, remain poorly understood. In this regard, we have recently demonstrated that LinCD34 and LinCD34+ cells have different functional characteristics regarding their kinetic status, clonogenic activity, and engraftment capacity [3].

Based on these considerations, in this study, we attempted to correlate the different functional properties of three subsets of hemopoietic stem/progenitor cells with their molecular phenotype. For this purpose, we used the DNA microarray technology to compare the expression profile of LinCD34, LinCD34+, and Lin+CD34+ stem/progenitor cells isolated from peripheral blood.

Our data showed a progressive increase of mRNA complexity from LinCD34 to LinCD34+ (+18.2%) cells and from LinCD34+ to Lin+CD34+ (+14.6%) cells, suggesting an increase of transcription activity.

In addition, the overall analysis of similarity, condition tree clustering, pairs the two CD34+ population transcript profiles, suggesting that the CD34 induction is associated with a strong variation of the global gene expression profile.

Functional analysis of differentially expressed gene categories confirmed that CD34 induction is correlated with cell proliferation, as previously suggested [3, 13], and is associated with a general metabolic activation, including DNA, RNA, and protein synthesis. The significant downregulation in CD34+ cells of pathways inhibiting HSCs proliferation, such as those of TGFβ, IL-10, and IL-17, may be responsible for the strong differential expression of cyclins, CDKs, CDK inhibitors, and cell cycle–related genes observed between CD34 and CD34+ cells.

Functional assays correlated the expression level of cytokine receptors and the clonogenic activity of LinCD34 and LinCD34+ cells. Interestingly, this is the first observation concerning the inhibitory effect of IL-17 on clonogenic activity of HSCs. These data give a strong molecular support to the different kinetic status of the three cell populations under study, indicating that LinCD34 cells are mainly in G0 phase, whereas Lin and Lin+CD34+ are cycling mainly in the G1 phase of the cell cycle, as already suggested by flow cytometry studies [3, 13].

In this study we confirm that CD34 induction in HSCs is correlated with cell cycle entry and lineage commitment. As far as self-renewal regulation is concerned, we found the preferential expression of several Hox genes and of GATA-2 in CD34+ cells. In particular, we showed that CD34+ cells coexpress HOXA9 and its cofactors Meis1 and PBX2, which are master regulators of HSC self-renewal [43]. The HOXB4 gene, a TPO-induced regulator of HSC self-renewal [28], is strongly upregulated in the CD34/CD34+ transition. Conversely, starting from our data, WNT and NOTCH pathways seem to be not relevant for human HSC self-renewal.

Acquisition of CD34 was also associated with the upregulation of transcription factors involved in the lineage commitment. In fact, the expression of genes implicated in erythroid, mega-karyocytic, and granulocytic commitment was strongly increased during the CD34/CD34+ transition.

As far as the engraftment capacity is concerned, the preferential expression in LinCD34+ of genes that positively regulate HSC homing and engraftment, such as VLA4, VLA5, and SEL-PLG [44, 45], provided the molecular basis for the higher engraftment capacity of CD34+ cells [3, 12].

Because CXCR4 was preferentially expressed in CD34 cells, their limited engraftment capacity can be ascribed to the concomitant higher expression of RGS1 and RGS13, two inhibitors of the CXCR4 signal transduction pathway.

VLA-4, VLA-5, and CXCR4, in fact, have a prominent role in the adhesion interactions involved in HSC homing and mobilization. Treatment of CD34+ cells with anti-VLA-4 or anti-VLA-5 prevented engraftment [45], whereas inhibition of the CXCR4-SDF1 pathway was associated with HSC mobilization [46]. Kollet et al. [47] have recently demonstrated that CD34+CXCR4 cells bear intracellular CXCR4, which can be functionally expressed to induce NOD/SCID repopulation. These data further underline our findings of dynamic CXCR4 expression on hematopoietic stem and progenitor cells [48].

Although LinCD34 cells displayed a preferential expression of nonhematopoietic epithelial, neural, and muscle markers, none of these genes can be considered a transcription regulator of tissue determination or differentiation. It is consequently difficult to assert that this expression pattern underlies the differentiation plasticity of HSCs [49]. In summary, we showed that the CD34 induction is strictly dependent on cell proliferation, because CD34 promoter is activated by cell cycle TFs such as NFY and C-myb, which, based on our results, are strongly increased during CD34/CD34+ transition. It is still unknown how the CD34 Ag is involved in the induction of differentiation that occurs in Lin+CD34+ cells, expressing a mixture of lineage-associated markers and transcription factors, as depicted in Figures 5A and 5C.

In general, models of stem cell regulation have been considered hierarchical, with a primitive HSC giving rise to proliferating progenitors and then to committed precursors [50]. Several reports in literature are not completely in agreement with the hierarchical model and have been recently reviewed [51, 52]. In fact, rather than the hierarchical transition from stem to progenitor cells, a more flexible system seems to be operative where the phenotype changes reversibly from HSC to progenitor and back depending on the kinetic status. The chromatin remodeling, associated with the cell cycle entry, causes the phenotypic changes on stem/progenitor cells and determinates the response to environmental stimuli.

Our data suggest that stem cell biology is largely dependent on their kinetic status. Most LinCD34 cells are quiescent, but with appropriate microenvironment stimuli can enter the cycle and proliferate, giving rise to Lin+CD34+ cells. Therefore, the chromatin remodeling occurring mainly in the G1 phase of the cycle, as demonstrated by the preferential expression of transcriptional activators in CD34+ context, probably underlies the transcriptome changes observed in Lin+CD34+ stem/progenitors cells.

The subsequent increasing transcriptional activity may be the premise for the activation of the genetic programs underlying self-renewal, commitment, and engraftment of HSCs.

Acknowledgements

R.M.L. and S.F. contributed equally to this study. This work was funded by MURST-COFIN 2002, Associazione Italiana per la Ricerca sul Cancro, and the University of Bologna (funds for selected topics).

Ancillary