We performed the first genome-wide expression analysis directly comparing the expression profile of highly enriched normal human hematopoietic stem cells (HSC) and leukemic stem cells (LSC) from patients with acute myeloid leukemia (AML). Comparing the expression signature of normal HSC to that of LSC, we identified 3,005 differentially expressed genes. Using 2 independent analyses, we identified multiple pathways that are aberrantly regulated in leukemic stem cells compared with normal HSC. Several pathways, including Wnt signaling, MAP Kinase signaling, and Adherens Junction, are well known for their role in cancer development and stem cell biology. Other pathways have not been previously implicated in the regulation of cancer stem cell functions, including Ribosome and T Cell Receptor Signaling pathway. This study demonstrates that combining global gene expression analysis with detailed annotated pathway resources applied to highly enriched normal and malignant stem cell populations, can yield an understanding of the critical pathways regulating cancer stem cells.
With the ability to enrich for rare populations of cells, using cell sorting techniques, and the development of appropriate xenotransplant models, it has been possible to prospectively characterize the surface antigen phenotype of both normal and leukemic stem cell populations from normal and AML specimens. A number of lines of evidence have demonstrated that human HSC are contained in the Lin-CD34+CD38- fraction of hematopoietic progenitors (1, 2). Additional studies have demonstrated that human HSC also express CD90 (35). Perhaps the best demonstration of HSC function comes from human clinical trials of autologous mobilized peripheral blood in clinical transplantation, where long-term engraftment was provided by transplantation of purified CD34+CD90+ cells (68). In multiple reports (1214), a common phenotype for AML LSC has been identified and found to be negative for expression of lineage markers (Lin-), positive for expression of CD34, and negative for expression of CD38 (911). These Lin-CD34+CD38- LSC were further shown to be positive for expression of IL-3Rα (CD123) and negative for expression of CD90 (Thy-1).
The cancer stem cell model has significant implications for the design of therapies for AML and other cancers. It postulates that to eradicate the tumor, therapies must target and eliminate the cancer stem cells. For the development of such cancer stem cell-targeted therapies, it is necessary to identify molecules and pathways that are preferentially expressed in these cancer stem cells compared with their normal counterparts.
DNA microarray technology has proven to be a powerful tool for the large scale analysis of gene expression differences, particularly in cancer investigations (15). There are many reports of gene expression profiles of bulk human AML samples, including several very large cohorts (1618). However, few studies have directly compared AML to normal hematopoietic cells (19). Comparisons of unfractionated populations may fail to identify critical differentially expressed genes and pathways within infrequent LSC and rare HSC populations. One report has investigated gene expression differences between LSC and non-LSC from the same AML samples to identify genes critical to LSC function; however, no direct comparison was made to normal HSC (20). An additional study has investigated gene expression differences between normal HSC and HSC from patients with 5q- myelodysplastic syndrome, and identified genes potentially involved in the pathogenesis of this disorder (21). We report the first analysis of gene expression differences between AML LSC and normal bone marrow HSC, using 2 independently derived datasets. We have applied the gene sets to a systems level pathway analysis and have identified molecular pathways and networks that are dysregulated between AML LSC and HSC.


Generation of Gene Expression Profiles of Human HSC and AML LSC.

Two independent sets of microarray gene expression profiles of human bone marrow HSC and AML LSC were generated at the University of Michigan (n = 3 HSC, n = 7 AML) and Stanford University (n = 4 HSC, n = 9 AML) by fluorescence-activated cell sorting (FACS) of patient samples, followed by RNA purification, amplification, and hybridization to Affymetrix oligonucleotide-based microarrays. The clinical features of the AML samples are presented in Table S1, and cover a range of subtypes of AML. These datasets were then combined for the bioinformatic analysis of differentially regulated pathways.

Dysregulated Pathways Between AML LSC and HSC.

Although LSC and normal HSC share the common characteristics of unlimited self-renewal and multilineage differentiation, understanding how they differ from each other should reveal fundamental mechanisms governing leukemic transformation. Although differentially expressed genes between LSC and HSC can be identified and validated individually, the full potential of genome-wide microarray analysis can be better realized in terms of gene regulatory networks, given that those genes, and the proteins they encode for, function in the context of intertwining pathways. We conducted an unbiased systems level pathway analysis without excluding any genes through a priori gene expression thresholds, by employing a newly developed algorithm dubbed integrative microarray analysis of pathways (IMAP) (22). This analysis first combines microarray data from multiple independent experiments through metaanalysis; a score for each pathway is assigned without specific gene expression cut-offs; the significance of each pathway is then computed by running 1 million iterations of randomly sampled genes. We used the pathway information from the Biocarta (www.biocarta.com), KEGG (www.genome.jp/kegg), GeneGo (www.genego.com), and Pathway Studio (www.ariadnegenomics.com/products/pathway-studio) databases as references to derive the top dysregulated pathways (see Fig. 1). As shown in Table 1, among the top dysregulated pathways between LSC and HSC, are pathways involved in adherens junction, regulation of the actin cytoskeleton, apoptosis, MAPK signaling, and Wnt signaling. A full list of the top dysregulated pathways derived from all 4 databases, and up- or down-regulated pathways, is shown in Table S2. Among the down-regulated pathways in LSC are those related to tumor suppressors (such as RB and ATM signaling), CXCR4→Stat3/5B pathways, and regulation of translation initiation. Thus, global pathway analysis has identified critical biological networks perturbed in LSC compared with normal HSC.
Fig. 1.
Schematic representation of dysregulated pathway identification.
Table 1.
Top 10 dysregulated pathways using the KEGG database
PathwaySourceGeneHitDys-P valueUp-P valueDown-P value
Adherens junctionKEGG847900.1402190.859781
Regulation of actin cytoskeletonKEGG2212140.0000020.0000150.999985
Tight junctionKEGG1331280.0000120.9594370.040563
Focal adhesionKEGG2392340.0000140.0007760.999224
MAPK signaling pathwayKEGG2512490.0002980.0172950.982705
T cell receptor signaling pathwayKEGG1041030.0006940.0212480.978752
Jak-STAT signaling pathwayKEGG1641620.0007530.0281080.971892
Wnt signaling pathwayKEGG1541490.0023040.5090850.490915
This table summarizes the top 10 dysregulated KEGG pathways according to their dys-regulation score (labeled dys-P value). Shown are the number of genes referenced in the KEGG pathways (gene) and the number of genes that were found in our data set (hit). For each KEGG pathway, an unbiased systems level pathway analysis without excluding any genes was implemented by employing integrative microarray analysis of pathways (IMAP). The P values of each pathway reflect the significance of dys-regulation (dys-P value), up-regulation (up-P value), and down-regulation (down-P value) of the pathway.

Identification of Key Molecular Interactions Within Dysregulated Pathways.

To identify changing molecular interaction and reaction networks contributing to the top dysregulated pathways between LSC and HSC, we mapped the relative expression levels of all of the genes found in a given pathway in the context of their signal transduction and cell communication processes as defined by the KEGG pathway database, using the Advanced Pathway Painter program (www.gsa-online.de/eng/app.html). As shown in Fig. 2, 2 of the top dysregulated pathways—adherens junction (2a) and Wnt signaling pathway (2b)—are illustrated with red color representing genes up-regulated in LSC, green color representing genes down-regulated in LSC, and yellow color representing genes with no significant change between LSC and HSC. Among the genes that are down-regulated in LSC in the adherens junction pathway are α-Catenin, Afadin, and PAR3 (Fig. 2A). Genes that are up-regulated in LSC in the Wnt pathway include Axin and APC, whereas c-Jun, a TCF/Lef target gene is down-regulated in LSC (Fig. 2B).
Fig. 2.
Visualization of molecular interaction and reaction networks in the KEGG database. (A) Adherens junction. (B) Wnt signaling pathway. Red color represents up-regulation in LSC, green color represents up-regulation in HSC, and yellow color represents no significant change.

Validation of Top Dysregulated Pathways, Using Functional Groups Enrichment Analysis.

Independent of the IMAP analysis, which takes into account all of the genes in a given pathway, we conducted a parallel analysis employing only differentially expressed genes (DEGs) from the combined datasets. The 2 AML microarray datasets were first combined as described above, using z scores. With P < 0.05 we obtained 3005 DEGs (Table S3). A heat map illustrating the expression of these genes across the Stanford samples is shown in Fig. S1. A functional enrichment analysis using this set of genes was performed using the DAVID program (the Database for Annotation, Visualization and Integrated Discovery; http://david.abcc.ncifcrf.gov/home.jsp) (23). The latest version of DAVID supports 40 annotation categories, including Gene Ontology terms, KEGG pathways, protein–protein interactions, protein functional domains, swissprot keywords, disease associations, bio-pathways, sequence general features, and literature. A number of functional groups or biological themes are identified as enriched in our AML dataset compared with the whole genome distribution of that specific group, using DAVID (P < 0.02). We developed a bioinformatic tool to visualize these enriched functional group associations, using the Cytoscape program (www.cytoscape.org) (24). As shown in Fig. 3, among the 3005 DEGs, genes involved in protein kinase activity, adherens junction, actin cytoskeleton, and apoptosis are specifically enriched. The size of each circle represents the number of genes in that specific functional group; the thickness of the lines represents the number of overlapping genes between the functional groups. Thus, 2 independent network analyses arrive at some common pathways differentially regulated between LSC and HSC, strongly implicating them in regulation of leukemic stem cell functions.
Fig. 3.
Functional Groups Association Networks (FGAN) visualized using the Cytoscape program. Each node represents 1 enriched functional group (GO molecular function, GO cellular component, and Swissprot keywords) with P value <0.02 performed by DAVID. A total of 3005 differentially expressed genes were evaluated by the functional enrichment analysis. The size of the node is proportional to the number of genes in each functional group. The largest functional group is the protein kinase activity of GO molecular function, which contains 103 genes. The edge width represents the number of shared genes between any two functional groups.


Although there is an abundance of data examining the gene expression profiles of normal and malignant bulk cell populations in AML, little has been done in the application of the stem cell model to gene expression analysis of AML samples. This is due in part to the difficulty in isolating a sufficient number of cells to perform these studies. Several studies have used CD34+ cells for microarray analyses, however, the expression of CD34 is often aberrant in AML and most studies to date have demonstrated there remains considerable heterogeneity in the CD34 positive fraction (9, 12, 13). We and others have previously demonstrated the value of applying macro- and microarray technology to highly enriched populations of normal stem cells and their committed progeny to identify key regulators of cell fate choices (2527). In 2001, Guzman et al. (28) applied macroarray technology to compare leukemic and normal stem cell populations and identified, among other genes, activation of the NFκB pathway in all of the LSC samples examined. Recently Gal et al. (20) examined the gene expression profile of enriched leukemic progenitors from 5 patients, using microarray technology, and compared the results of their study to a dataset previously published for normal hematopoietic progenitors. The current study represents the first effort to simultaneously compare the transcriptional profiles of highly enriched LSC and normal HSC from a wide variety of patients, using modern microarray technology. The direct comparison of expression patterns of LSC to HSC, as opposed to other non-stem cell populations, enhanced our ability to identify genes and pathways that are disrupted at the stem cell level in AML. This analysis provides critical insights into the differences between normal and malignant stem cell populations that may be used for the development of targeted therapies, and tools for assessing the impact of therapy on the LSC population.
The first dataset was generated from sorted leukemic stem cells (LSC) from 7 AML patients and normal hematopoietic stem cells (HSC) from 3 normal controls. The second dataset was generated similarly from 9 AML patients and 4 normal individuals. We obtained 3,005 differentially expressed genes with 1,451 genes being up-regulated and 1,554 genes being down-regulated in LSC when compared with normal HSC. Many of these genes have been previously identified as being playing a key role in normal stem cell biology and in leukemia. Despite the use of different microarray platforms, a detailed analysis of the 2 independently derived datasets demonstrated a significant degree of overlap between the gene signatures for the 2 cell types examined.
In addition to providing a genome-wide survey for genes whose expression is disrupted in leukemic transformation at the level of the stem cell compartment, a comprehensive, unbiased pathway analysis has allowed the identification of critical pathways which are dysregulated in LSC compared with normal HSC. Again, focusing on the difference between normal and leukemic as opposed to stem cell versus progenitor has allowed us to screen out many of the pathways that are involved in stem cell function which are not dysregulated in the leukemic state. Hence, the limited number of pathways we identified as being dysregulated. Furthermore, the significant degree of overlap within and between the different pathway tools used suggests that these pathways are critical in the evolution of cancer stem cells from their normal counterparts.
To validate the datasets used, we performed an analysis employing only the 3,005 differentially expressed genes (DEGs) from the combined datasets, using the DAVID program. This independent and parallel analysis demonstrated a significant degree of overlap with the unbiased pathway analysis and confirmed the validity of the datasets used in these analyses, using a system-wide approach.
Many of the pathways we identified as being aberrantly regulated in the LSC from the patients studied have already been established as playing key roles in leukemia and/or leukemic stem cell biology including the Wnt canonical (29), the Adherens junction and NFκB pathways (28). Several of the pathways identified are involved in the interaction of the stem cells with their niche. There is a growing body of data demonstrating the importance of the interaction of stem cells with their niche in normal and malignant stem cell biology. Two classes of receptors and their ligands are critical in determining the nature of this interaction, cell adhesion molecules (CAMs) and chemokine receptors. Our current analysis demonstrates dysregulation of both gene families in the leukemic stem cells studied. Identification of dysregulation of these interactions is consistent with the hypothesis that alteration of the stem cell:niche interaction is a key step in the pathogenesis of cancer stem cells (30). In regard to CAMs, we identified several CAM related pathways that were aberrantly regulated in LSC including the Adherens junction and Tight junction pathways. In addition, pathway analysis of the datasets, using the Biocarta, GeneGo, and Pathway Studio tools, identified: How Salmonella hijacks a cell, Adhesion Molecules on Lymphocyte, CXCR4 → STAT3 signaling pathway, CXCR4 → STAT5B signaling pathway, and the Angiopoietin–Tie2 signaling pathways (Table S2).
The adherens junction has been demonstrated to be critical in the interaction of HSC and niche in both the fetal liver stage and adult stage of hematopoiesis (3133). Our data confirms that N-cadherin and alpha-E catenin are expressed in normal human stem cells and that expression of these genes is disrupted in LSC. Several studies have also demonstrated that elements of the adherens junction are aberrantly expressed in leukemic cells compared with normal hematopoietic cells (3438). We recently identified aberrant expression of CTNNA1 in highly enriched LSC from patients with advanced MDS and AML associated with abnormalities of chromosome 5 (39). Likewise, the Angiopoietin–Tie2 signaling pathway has been demonstrated to play an important role in normal and leukemic stem cell function (35, 36), and expression of Tie2 has been shown to be lost in LSC. Finally, the selectins, and members of the tight junction complex are also involved in the interaction of normal stem cells with endothelial cells and are dysregulated during the leukemic transformation process.
Another family of proteins critical to the normal and leukemic stem cell: Niche interaction is the chemokine family. CXCR4 is a chemokine receptor that plays a key role in regulating normal and leukemic stem cell homing to the bone marrow niche. The role of CXCR4 signaling in AML is of significant interest as several targeted therapies that disrupt the interaction of CXCR4 with its ligand CXCL12 are currently being investigated in clinical trials.
The Wnt canonical pathway was found to be dysregulated in this study between LSC and HSC, and has been implicated in the pathogenesis of several different types of human cancer, including leukemia. We have previously shown that the canonical Wnt pathway, signaling through nuclear beta-catenin, regulates the self-renewal of mouse HSC (40, 41). We have also shown that this pathway is aberrantly activated in downstream progenitors in the blast crisis phase of chronic myelogenous leukemia, resulting in the nuclear localization of beta-catenin where it likely acts to stimulate self-renewal and contributes to the formation of leukemia stem cells (29). Our data suggests that this pathway may be dysregulated in AML stem cells, possibly contributing to pathogenesis.
Many pathways not previously implicated in the regulation of leukemia stem cell functions were identified in our analysis. Several basic cellular biology pathways identified include Ribosome, Regulation of Actin Cytoskeleton, and Regulation of Translation Initiation. Numerous metabolic pathways were identified such as Glycosphingolipid, Androgen and Estrogen, Glycerophospholipid, Regulation of Fatty Acid Synthase Activity, and Arginine Metabolism. Other pathways involved signal transduction including Angiotensin II, Oxidative Stress-Induced Gene Expression, T Cell Receptor, CD28, B Cell Receptor, and EGF Signaling. Ultimately, detailed biological investigations will be necessary to determine the functional involvement of these pathways in leukemic pathogenesis.
In summary, we have used gene expression profiling and annotated pathway resources to identify biological networks that are dysregulated in AML stem cells compared with normal HSC. The application of the stem cell model for AML to a systems biology analysis of stem cell expression networks has confirmed the role of several pathways previously demonstrated to be important in cancer stem cell function. In addition, this approach has identified several pathways not previously studied in cancer stem cells. Such networks are candidates for involvement in the regulation of critical cancer stem cell functions, and as such may be targets for therapeutic intervention.

Materials and Methods

Human Samples.

Normal human bone marrow mononuclear cells were either purchased from AllCells or obtained from National Marrow Donor Program discarded filter units. For AML specimens, peripheral blood and/or bone marrow was obtained following informed consent at the time of clinical presentation according to IRB approved protocols at the University of Michigan or Stanford University. Mononuclear cells were prepared using Ficoll-Paque Plus (GE Healthcare), and either used fresh or cryopreserved in 90% FBS/10% DMSO in liquid nitrogen.

Isolation and Purification of Normal HSC and AML LSC.

Viably frozen cells were thawed and resuspended in IMDM with 2% heat-inactivated FBS and DNase I (Sigma). For all normal specimens and leukemic samples with <10% of blasts expressing CD34, enrichment was performed before staining, using CD34 positive selection (Stem Cell Technologies, Canada and Miltenyi Biotech, Germany). Cells were stained as described in ref. 39 (SI Methods) and analyzed and sorted using FACSAria cytometers (BD Biosciences). A total of 15,000–65,000 normal HSC and ≈50,000–150,000 AML LSC were sorted for RNA purification.

RNA Purification, Amplification, and Microarray Analysis.

Total RNA was extracted using TRIzol reagent containing either glycogen or linear acrylamide according to the manufacturer's protocol, and then treated with DNaseI (Ambion). For the University of Michigan samples only, RNA was reextracted in TRIzol. All RNA samples were quantified with the RiboGreen RNA Quantitation Kit (Molecular Probes), subjected to reverse transcription, 2 consecutive rounds of linear amplification, and production and fragmentation of biotinylated cRNA (Affymetrix). 15 μg of cRNA from each sample was hybridized to Affymetrix HG U133 A or B (University of Michigan samples) or HG U133 Plus 2.0 (Stanford University samples) microarrays. Hybridization and scanning were performed according to the manufacturer's instructions (Affymetrix). The data from the University of Michigan and Stanford expression array studies was uploaded into ISB's Systems Biology Experiment Analysis Management System (SBEAMS) and normalized using the GCRMA algorithm.

Integration of Dataset.

Because the 2 datasets of LSC and HSC were obtained from different platforms (3 different chips) and/or different probe sets, the ability to directly compare the datasets to each other was limited, and the correlation coefficient between the 2 datasets was low. To obtain a statistically more robust dataset, we developed a strategy to combine both datasets for an increased sample size (Fig. 1).
We performed metaanalysis for integration of the 2 different microarray datasets, using a procedure similar to that used in Setlur et al. (22). P values were first calculated for all genes within each dataset using the Wilcoxon rank-sum test to compare LSC against HSC samples. These P values were then mapped onto a standard normal curve. That is, z scores corresponding to the P values were obtained
where p is the P value to be standardized, μ the mean of all P values within a dataset, and σ the standard deviation of the P values within the dataset. After standardization, the 2 datasets were integrated to get a weighted z score for each gene by combining all z scores for the gene across the 2 datasets, using the following Liptak–Stouffer formula:
where m = 2 for this study, i.e., the number of datasets; wi = 1, the weight of the ith dataset; zi,g the z score of gene g in the ith dataset. The less the value of zcomb,g, the less the probability of differential expression between HSC and LSC by chance. The probability can be calculated from the cumulative standard normal distribution:
Based on these probabilities, namely P values, we could integrate our data without bias, and finally obtained 3,005 potential differentially expressed genes with P values <0.05, and 387 genes with P values <0.01.

Dysregulation of Pathway.

To define the dysregulation of a pathway P, we first assigned a score sg to each gene g in the pathway P by the negative logarithm of its P value pg, which was
We then gave a total score SP to the pathway P by summing up all scores of genes in the pathway, namely
To estimate a P value for significance of this pathway, we iteratively computed similar scores 1 million times on randomly generated pathways of the same size as that of pathway P. The frequency of scores that were larger than SP was used as the P value of pathway P to describe its dysregulation. For calculation of up- or down-regulated pathways, a positive or negative sg value was used respectively to compute the Sp value.


We thank Libuse Jerabek for excellent laboratory management. This work was supported by the Walter and Idun Y. Berry Foundation, the Doctors Cancer Foundation, and the AACR Postdoctoral Fellowship in Cancer Research (R.M.) and National Institutes of Health Grants R01CA86017 (to I.L.W.), 5K08CA100138 (to M.W.B.), and P01DK53074 (to M.F.C., I.L.W., and L.H.). R.M. holds a Career Award for Medical Scientists from the Burroughs Wellcome Fund.

Supporting Information

Table S1 (PDF)
Supporting Information
Table S2 (PDF)
Supporting Information
Supporting Information (PDF)
Supporting Information


M Bhatia, JC Wang, U Kapp, D Bonnet, JE Dick, Purification of primitive human hematopoietic cells capable of repopulating immune-deficient mice. Proc Natl Acad Sci USA 94, 5320–5325 (1997).
M Kondo, et al., Biology of hematopoietic stem cells and progenitors: Implications for clinical application. Annu Rev Immunol 21, 759–806 (2003).
CM Baum, IL Weissman, AS Tsukamoto, AM Buckle, B Peault, Isolation of a candidate human hematopoietic stem-cell population. Proc Natl Acad Sci USA 89, 2804–2808 (1992).
W Craig, R Kay, RL Cutler, PM Lansdorp, Expression of Thy-1 on human hematopoietic progenitor cells. J Exp Med 177, 1331–1342 (1993).
L Murray, et al., Enrichment of human hematopoietic stem cell activity in the CD34+Thy-1+Lin- subpopulation from mobilized peripheral blood. Blood 85, 368–378 (1995).
M Michallet, et al., Transplantation with selected autologous peripheral blood CD34+Thy1+ hematopoietic stem cells (HSCs) in multiple myeloma: Impact of HSC dose on engraftment, safety, and immune reconstitution. Exp Hematol 28, 858–870 (2000).
RS Negrin, K Atkinson, T Leemhuis, E Hanania, C Juttner, et al., Transplantation of highly purified CD34+Thy-1+ hematopoietic stem cells in patients with metastatic breast cancer. Biol Blood Marrow Transplant 6, 262–271 (2000).
JM Vose, et al., Transplantation of highly purified CD34+Thy-1+ hematopoietic stem cells in patients with recurrent indolent non-Hodgkin's lymphoma. Biol Blood Marrow Transplant 7, 680–687 (2001).
D Bonnet, JE Dick, Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med 3, 730–737 (1997).
T Lapidot, et al., A cell initiating human acute myeloid leukaemia after transplantation into SCID mice. Nature 367, 645–648 (1994).
JC Wang, JE Dick, Cancer stem cells: Lessons from leukemia. Trends Cell Biol 15, 494–501 (2005).
A Blair, DE Hogge, LE Ailles, PM Lansdorp, HJ Sutherland, Lack of expression of Thy-1 (CD90) on acute myeloid leukemia cells with long-term proliferative ability in vitro and in vivo. Blood 89, 3104–3112 (1997).
CT Jordan, et al., The interleukin-3 receptor alpha chain is a unique marker for human acute myelogenous leukemia stem cells. Leukemia 14, 1777–1784 (2000).
T Miyamoto, IL Weissman, K Akashi, AML1/ETO-expressing nonleukemic stem cells in acute myelogenous leukemia with 8;21 chromosomal translocation. Proc Natl Acad Sci USA 97, 7521–7526 (2000).
J Quackenbush, Microarray analysis and tumor classification. N Engl J Med 354, 2463–2472 (2006).
L Bullinger, et al., Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med 350, 1605–1616 (2004).
L Bullinger, PJ Valk, Gene expression profiling in acute myeloid leukemia. J Clin Oncol 23, 6296–6305 (2005).
PJ Valk, et al., Prognostically useful gene-expression profiles in acute myeloid leukemia. N Engl J Med 350, 1617–1628 (2004).
DL Stirewalt, et al., Identification of genes with abnormal expression changes in acute myeloid leukemia. Genes Chromosomes Cancer 47, 8–20 (2008).
H Gal, et al., Gene expression profiles of AML derived stem cells; similarity to hematopoietic stem cells. Leukemia 20, 2147–2154 (2006).
L Nilsson, et al., The molecular signature of MDS stem cells supports a stem-cell origin of 5q myelodysplastic syndromes. Blood 110, 3005–3014 (2007).
SR Setlur, et al., Integrative microarray analysis of pathways dysregulated in metastatic prostate cancer. Cancer Res 67, 10296–10303 (2007).
G Dennis, et al., DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 4, P3 (2003).
P Shannon, et al., Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res 13, 2498–2504 (2003).
EC Forsberg, et al., Differential expression of novel potential regulators in hematopoietic stem cells. PLoS Genet 1, e28 (2005).
NB Ivanova, et al., A stem cell molecular signature. Science 298, 601–604 (2002).
IK Park, et al., Differential gene expression profiling of adult murine hematopoietic stem cells. Blood 99, 488–498 (2002).
ML Guzman, et al., Nuclear factor-kappaB is constitutively activated in primitive human acute myelogenous leukemia cells. Blood 98, 2301–2307 (2001).
CH Jamieson, et al., Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML. N Engl J Med 351, 657–667 (2004).
MF Clarke, M Fuller, Stem cells and cancer: Two faces of eve. Cell 124, 1111–1115 (2006).
LM Calvi, et al., Osteoblastic cells regulate the haematopoietic stem cell niche. Nature 425, 841–846 (2003).
I Kim, OH Yilmaz, SJ Morrison, CD144 (VE-cadherin) is transiently expressed by fetal liver hematopoietic stem cells. Blood 106, 903–905 (2005).
J Zhang, et al., Identification of the haematopoietic stem cell niche and control of the niche size. Nature 425, 836–841 (2003).
G Acs, VA LiVolsi, Loss of membrane expression of E-cadherin in leukemic erythroblasts. Arch Pathol Lab Med 125, 198–201 (2001).
F Arai, et al., Tie2/angiopoietin-1 signaling regulates hematopoietic stem cell quiescence in the bone marrow niche. Cell 118, 149–161 (2004).
C Schliemann, et al., Expression of angiopoietins and their receptor Tie2 in the bone marrow of patients with acute myeloid leukemia. Haematologica 91, 1203–1211 (2006).
T Shimamoto, JH Ohyashiki, K Ohyashiki, Methylation of p15(INK4b) and E-cadherin genes is independently correlated with poor prognosis in acute myeloid leukemia. Leuk Res 29, 653–659 (2005).
L Wang, H O'Leary, J Fortney, LF Gibson, Ph+/VE-cadherin+ identifies a stem cell like population of acute lymphoblastic leukemia sustained by bone marrow niche cells. Blood 110, 3334–3344 (2007).
TX Liu, et al., Chromosome 5q deletion and epigenetic suppression of the gene encoding alpha-catenin (CTNNA1) in myeloid cell transformation. Nat Med 13, 78–83 (2007).
T Reya, et al., A role for Wnt signalling in self-renewal of haematopoietic stem cells. Nature 423, 409–414 (2003).
K Willert, et al., Wnt proteins are lipid-modified and can act as stem cell growth factors. Nature 423, 448–452 (2003).

Information & Authors


Published in

Go to Proceedings of the National Academy of Sciences
Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 106 | No. 9
March 3, 2009
PubMed: 19218430


Submission history

Received: April 6, 2008
Published online: March 3, 2009
Published in issue: March 3, 2009


We thank Libuse Jerabek for excellent laboratory management. This work was supported by the Walter and Idun Y. Berry Foundation, the Doctors Cancer Foundation, and the AACR Postdoctoral Fellowship in Cancer Research (R.M.) and National Institutes of Health Grants R01CA86017 (to I.L.W.), 5K08CA100138 (to M.W.B.), and P01DK53074 (to M.F.C., I.L.W., and L.H.). R.M. holds a Career Award for Medical Scientists from the Burroughs Wellcome Fund.


This article contains supporting information online at www.pnas.org/cgi/content/full/0900089106/DCSupplemental.



Ravindra Majeti1
Institute for Stem Cell Biology and Regenerative Medicine and Department of Internal Medicine, Divisions of Hematology and Oncology, Stanford University, Palo Alto, CA 94304;
Michael W. Becker1
Division of Hematology/Oncology, University of Rochester, Rochester, NY 14627;
Qiang Tian1
Institute for Systems Biology, Seattle, WA 98103;
Department of Medicine/Hematology, University of Washington School of Medicine, and the Institute for Stem Cell and Regenerative Medicine, Seattle, WA 98103;
Tsung-Lu Michael Lee
Institute for Systems Biology, Seattle, WA 98103;
Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan 70101, Taiwan; and
Xiaowei Yan
Institute for Systems Biology, Seattle, WA 98103;
Rui Liu
University of Michigan School of Dentistry, Ann Arbor, MI 48109
Jung-Hsien Chiang
Institute for Systems Biology, Seattle, WA 98103;
Department of Computer Science and Information Engineering, National Cheng Kung University, Tainan 70101, Taiwan; and
Leroy Hood
Institute for Systems Biology, Seattle, WA 98103;
Michael F. Clarke
Institute for Stem Cell Biology and Regenerative Medicine and Department of Internal Medicine, Divisions of Hematology and Oncology, Stanford University, Palo Alto, CA 94304;
Irving L. Weissman2 [email protected]
Institute for Stem Cell Biology and Regenerative Medicine and Department of Internal Medicine, Divisions of Hematology and Oncology, Stanford University, Palo Alto, CA 94304;


To whom correspondence should be addressed. E-mail: [email protected]
Contributed by Irving L. Weissman, January 6, 2009
Author contributions: R.M., M.W.B., Q.T., L.E.H., M.F.C., and I.L.W. designed research; R.M., M.W.B., Q.T., and R.L. performed research; Q.T., T.-L.M.L., X.Y., and J.-H.C. contributed new reagents/analytic tools; R.M., M.W.B., and Q.T. analyzed data; and R.M., M.W.B., Q.T., L.E.H., M.F.C., and I.L.W. wrote the paper.
R.M., M.W.B., and Q.T. contributed equally to this work.

Competing Interests

Conflict of interest statement: I.L.W. was a member of the scientific advisory board of Amgen and owns significant Amgen stock. I.L.W. cofounded and consulted for Systemix, is a cofounder and director of Stem Cells Inc., and cofounded Cellerant, Inc. L.H. is a scientific founder of Integrated Diagnostics, a protein diagnostics company using blood, and Amgen, a large biotechnology company.

Metrics & Citations


Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.

Citation statements



If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

Cited by


    View Options

    View options

    PDF format

    Download this article as a PDF file


    Get Access

    Login options

    Check if you have access through your login credentials or your institution to get full access on this article.

    Personal login Institutional Login

    Recommend to a librarian

    Recommend PNAS to a Librarian

    Purchase options

    Purchase this article to get full access to it.

    Single Article Purchase

    Dysregulated gene expression networks in human acute myelogenous leukemia stem cells
    Proceedings of the National Academy of Sciences
    • Vol. 106
    • No. 9
    • pp. 2969-3639







    Share article link

    Share on social media