Cytokine-Induced Mobilization of HSC
In adult mice and humans, the majority of HSC are found in the BM, although HSC are also constitutively present at low levels in the circulation 
. The frequency of HSC in the blood can be significantly increased through the use of mobilizing agents, including cytotoxic drugs and/or cytokines, which act to both drive HSC proliferation and to induce HSC migration from the BM into the bloodstream. In particular, treatment of mice with a combination of the chemotoxin cyclophosphamide (Cy) - which kills mainly proliferating hematopoietic progenitors and very few of the G0
HSC - plus the cytokine granulocyte-colony stimulating factor (G-CSF or G) induces a rapid and reproducible expansion and migration of BM HSC 
. Following administration of Cy + 2 daily doses of G, the BM HSC population (referred to hereafter as mobHSC) expands dramatically, reaching ~10–12 times the size of the HSC compartment in normal animals 
. Expansion of HSC in this early phase of mobilization occurs only in the BM 
, and appears to be associated with an increase in HSC self-renewal and an accelerated rate of HSC division, as measured by incorporation into newly synthesized DNA of the thymidine analog bromodeoxyuridine (BrdU) 
. While BM HSC of unmanipulated animals are largely quiescent (<10% of cells in S-G2
/M phases of the cell cycle), up to 35% of mobHSC exhibit >2n DNA content, and ~85% of these cells incorporate BrdU after only 12 hours of labeling 
. After day +2, HSC frequencies in the BM decline as HSC migrate from the marrow and begin to appear in significant numbers in the blood and spleen of mobilized animals 
To identify specific molecular mediators of the proliferative and migratory responses of HSC to Cy/G-induced hematopoietic stress, we used DNA microarray technology to identify changes in gene expression associated with HSC mobilization. This analysis yielded a dataset comprised of 2,611 targets exhibiting differential expression (); 2,461 clones (representing 1,670 individual genes) were significantly downregulated in mobHSC as compared to untreated BM HSC (Supplemental Table S1
), while only 150 clones (representing 97 individual genes) (Supplemental Table S2
) were significantly upregulated in response to Cy/G treatment. This bias towards reduced gene expression upon mobilization may suggest that downregulation of genes actively maintaining HSC quiescence is an important mechanism of cytokine-induced HSC expansion.
Significance Analysis of Microarray (SAM) plots.
Functional classification of genes differentially expressed by non-mobilized versus mobHSC revealed that ~half of these encode unknown genes or uncharacterized EST sequences. Gene Ontology (GO) analysis 
was performed to functionally classify the known genes, and revealed significant enrichment among up-regulated genes of cell cycle regulators, translation and RNA processing factors, and genes involved in basic cell metabolic processes. These findings are consistent with the robust activation of proliferation and uniform exit from quiescence induced in HSC by Cy/G treatment. Among the significantly down-regulated genes, a sizable proportion encoded transcription factors, signaling proteins, and cell cycle associated proteins, largely inhibitors of cell cycle progression (Supplemental Table S1
Leukemic Transformation of HSC
Leukemia is an aberrant hematopoietic process that can be initiated and sustained by rare leukemic stem cells (LSC) that drive the formation and growth of tumors 
. Recent evidence indicates that LSC can emerge either from transformed HSC 
or from transformed progenitor cells that have re-acquired the stem cell property of self-renewal 
. In one model, preleukemic progression of a myeloproliferative disorder occurs in the only self-renewing cells of the myeloid lineage, HSC, but emerges from these clones when self-renewal is not shut down [or re-emerges] when these preleukemic or chronic phase leukemic HSC develop multipotent or oligolineage progenitors at acute or blast crisis phases of the disease 
. Developing interventions that specifically target LSC is an appealing strategy for improving the specificity and efficiency of cancer treatment 
. Such targeting will require understanding of how LSC escape normal regulatory mechanisms to become malignant. Myeloid malignancies provide an excellent opportunity for addressing these fundamental questions at the cellular and molecular level as relevant LSC populations have been identified and can be distinguished and isolated apart from other cells in the tumor 
Mice lacking the AP-1 transcription factor JunB in hematopoietic cells develop a myeloproliferative disorder (MPD) that accurately reproduces important clinical aspects of human leukemias including chronic myelogenous leukemia (CML) and chronic myelomonocytic leukemia (CMML). We have identified the LSC population in these mice as arising from the HSC compartment 
. Loss of JunB function in HSC causes an aberrant stem cell expansion leading to MPD development and eventually frank leukemia. Inactivation of jun
B endows HSC with proliferative and survival advantages, two key properties necessary for HSC transformation into LSC.
To identify molecular mediators of leukemic transformation in junB-deficient HSC, we performed gene expression microarray analysis. Because gene expression profiles of LSC may differ depending on the stage of the disease, we investigated the LSC population of each leukemic mouse separately and selected for analysis only those mice exhibiting advanced MPD but no blast crisis progression. This strategy was designed to uncover genes involved in the initial steps of HSC transformation caused by loss of JunB expression, which in turn mediates the early phase of leukemia development.
RNA was isolated and amplified from 4,000–8,000 double-sorted HSC from the BM of individual jun
B-deficient mice. Target transcripts significantly up- or down-regulated were identified by Statistical Analysis of Microarrays (SAM) 
of six independent arrays according to the criteria described previously 
. These analyses detected 1,203 targets, representing 958 individual genes, that were significantly downregulated in jun
B-deficient HSC ( and Supplemental Table S3
). In contrast, only 26 target cDNAs, representing 23 individual genes, were upregulated in the same population (Supplemental Table S4
). This bias towards reduced gene expression suggests that JunB acts primarily as a transcriptional activator in HSC and suggests that downregulation of genes actively maintaining HSC function is an important prerequisite for their leukemic transformation.
Meta-Analysis Reveals a Normal HSC Signature
We previously employed gene expression profiling to study changes in gene regulation associated with the differentiation of self-renewing HSC to MPP 
, uncovering many novel pathways that regulate normal HSC function and specify HSC differentiation. Thus, together with the mobilization and leukemogenesis arrays described above, we have performed 3 pairwise comparisons to identify changes of gene expression associated with the differentiation, mobilization, or transformation of purified HSC (
; ). A heatmap representation of these comparisons is presented in . These three distinct processes have in common some characteristic cellular responses, for example the exit of HSC from quiescence and entry of these cells into cycle. To determine whether common molecular pathways mediate such shared HSC responses during differentiation, mobilization, and leukemic transformation, we next performed a meta-analysis of the differentially regulated genes from these three pairwise comparisons. We searched for significant two-way and three-way overlaps among the differentially expressed gene lists identified in the HSC vs. MPP, mobilized (Mob) vs. non-mobilized (nonMob), and leukemic (LSC) vs. non-leukemic (nonLSC) HSC datasets (). The significance of all two-way and three-way intersections was estimated using the hypergeometric distribution. Several of these intersections yielded no or few overlapping genes. In particular, given that both LSC (“LSC”, green, solid,) and mobHSC (“Mob”, blue, solid) exhibit substantial expansion of the stem cell population, we were surprised to find no overlap of significantly enriched genes in these populations. This result suggests that self-renewing divisions of mobilized HSC induced in response to Cy/G treatment are driven by pathways distinct from those that drive leukemic expansion upon loss of JunB.
Heatmap representation of differential gene expression.
Venn diagram representation of differential gene expression.
In contrast, intersection of genelists describing targets enriched in MPP relative to HSC, and in non-mobilized and non-leukemic HSC, did yield a short list of 32 clones, representing 30 unique targets. This set represents genes whose expression is increased as HSC differentiate to MPP, but reduced when HSC are mobilized or transformed. Interestingly, this list is highly enriched in nuclear proteins with transcriptional regulatory activity, including Esr1, Fos, Foxp1, Iep5, Laf4, and several zinc finger containing proteins: Zfp148, Zfp36, and Zfp469. A common property of MPP, mobilized HSC and LSC is their enhanced proliferative activity; however, unlike mobilized HSC and LSC, MPP harbor little or no self-renewal activity. Thus, the genes included in this set may represent those whose expression in proliferating cells is incompatible with self-renewal, such that they must be down-regulated to support expansion of self-renewing HSC (mobilization) or LSC (transformation), and are increased in MPP to promote differentiation.
Importantly, intersecting genes selectively enriched in non-mobilized and non-leukemic HSC with genes enriched in HSC when compared to MPP yielded 129 clones representing 93 individual genes ( and Supplemental Table S5
). The 93 unique genes in this intersection thus are expressed by “normal”, quiescent HSC, and are downregulated during differentiation, mobilization and leukemogenesis. This overlapping dataset is of particular interest as it likely includes genes that are crucial for the steady-state maintenance and regulation of a normal, slowly proliferating HSC.
Genes Enriched in Normal HSC
Using quantitative RT-PCR (qRT-PCR), we validated the differential expression of several of the 93 genes from the 3-way intersect described above using new, independently generated samples (). All genes in the HSC/MPP and HSC/mobHSC data sets showed the same trend by qRT-PCR as seen in the array analysis. While this concordance of data was also true for the vast majority of genes in the HSC vs. LSC comparison, some genes identified as downregulated by array analysis did not show significant downregulation when analyzed by qRT-PCR in an independent set of sorted samples. This variability likely relates to subtle differences in the precise stage of MPD at which individual junB-deficient animals were sacrificed, reflecting mouse-to-mouse variability in disease progression. Future studies to generate a more refined time course of differential gene expression in junB-deficient HSC may identify the order in which genes are up- and downregulated during leukemogenesis, and their interdependence or importance for disease initiation versus progression.
Quantitative RT-PCR of HSC, MPP, D+2 mobilized HSC and LSC.
We broadly categorized the 93 HSC “signature” genes based on GO analysis () and subcellular location. Enriched expression was observed for several genes encoding extracellular proteins, including biglycan, transglutaminase2, follistatin-like1, angiopoietin1 and procollagen4a2. Expression of several of these in “normal” HSC likely has significant implications for regulation of HSC quiescence. For example, biglycan is an extracellular matrix (ECM) proteoglycan that binds TGF-beta, collagen, and other ECM components. Interestingly, biglycan-deficient mice develop age-associated osteopenia thought to arise from defects in the maintenance and metabolic activity of osteogenic precursor cells 
. Biglycan deficiency also causes reduced responsiveness to TGFbeta among osteolineage cells 
. As TGF-beta has been reported as a negative regulator of HSC number, down-regulation of biglycan among differentiating, mobilizing and transformed HSC may insulate them from TGF-beta activity, thus facilitating their enhanced proliferative activity. Similarly, binding of angiopoietin 1 (Angpt1), a secreted cytokine, to its receptor Tie2 on HSC is reported to enhance HSC quiescence and to modulate expression of adhesion molecules implicated in HSC mobilization and retention in the BM 
. Interestingly, although previous studies have suggested that osteoblasts are the primary producers of Angpt1 in BM, our gene expression data suggest that HSC quiescence can also be modulated by autocrine effects.
Gene ontology classification of genes selectively expressed by normal, quiescent HSC.
Several membrane proteins, including Syndecan2, Edg1, calsyntenin1, Map17, Phemx, Ryk, Itm2a, Jam1 and Jam3, Robo4, Cish, Protein C receptor (PRCR), Prion protein, Mllt4 and Ica1 were enriched in the HSC signature. Several of these receptors are involved in cell migration or location in other systems. Edg1 encodes a receptor for sphingosine 1-phosphate (S1P). S1P recently has been characterized as a unique bimodal regulator of cell migration in the thymus and lymphatic system 
, an important determinant of plasma cell homing from lymph nodes to BM 
, and a key regulator of the physiologic recirculation of HSC and progenitor cells 
. Thus, downregulation of S1P receptor on HSC may facilitate their egress from the BM niche. Likewise, Jam2 is a leukocyte migration receptor that mediates heterotypic cell-cell interactions with integrin alpha(4)beta(1) (VLA-4), an important receptor in the BM homing and mobilization of HSC 
. Jam1 and Jam3 were recently shown to be reliable markers for HSC 
. Thus, the Jam family of adhesion molecules are likely important for HSC retention in the niche. Cish is a membrane protein involved in the downregulation of cytokine responsiveness, potentially acting as a growth inhibitor. Notably, the genomic region encoding Cish is frequently deleted in lung and kidney tumors, implicating this gene in tumor progression and suggesting that its loss in LSC might potentiate leukemic transformation of HSC 
. Ryk is a tyrosine-kinase-like receptor and a coreceptor for the Wnt signaling protein, previously implicated in the maintenance of mouse HSC 
. Ryk was shown recently to play roles in neural stem cell differentiation by Wnt-induced translocation to the nucleus 
. Finally, Robo4 is a homolog of the brain-specific Roundabout receptors that bind the secreted Slit ligands and play roles in axon guidance, while PRCR selectively marks HSC 
and Prion protein has been implicated in HSC self-renewal 
We also observed significant representation in the HSC gene set signature of genes involved in basic metabolic processes, including members of the glycogen phosphorylase (Pygl), aldehyde dehydrogenase (Aldh1a7), guanylate cyclase 1 (Gucy1a3), lactate dehydrogenase (Ldh2), and cytochrome P450 (Cyp2j6) families. Enrichment of these genes may reflect a differential requirement for certain metabolic effectors in the transition from a largely quiescent and metabolically inactive HSC to a highly proliferative and metabolically active cell upon differentiation, mobilization or transformation.
Nuclear proteins enriched among HSC signature genes include cell cycle regulators such as Cdkn1c, Necdin, and Rbbp9, and transcription factors such as Creb3l2, Mllt3, Pbx1, Prdm16, Smarca2 and Zfp467. Creb3l2 belongs to the basic leucine-zipper family of transcription factors and was identified as a balanced translocation fusion partner in low-grade fibromyxoid sarcoma 
. Ndrg2 is a candidate tumor suppressor gene on chromosome 14q that is inactivated during meningioma progression. Ndrg2 expression is induced by WT1, a transcriptional regulator frequently expressed in leukemias 
. Mllt3 (AF9) plays roles in hematopoietic differentiation into megakaryocyte and erythroid lineages 
and as a translocation partner with Mll causes leukemia via a leukemic stem cell mechanism. Thus, Mllt3 plays roles in normal cell growth and maintenance and in oncogenesis 
. Pbx1 cooperates with Hox proteins and has well-established roles in hematopoiesis 
. Intriguingly, it was recently suggested that Pbx1 promotes HSC quiescence and self-renewal 
. A short form of Prdm16 (MEL1) has been shown to block G-CSF-induced myeloid differentiation 
and thus may help to maintain the undifferentiated state of HSC, whereas Zfp467 augments and transactivates Stat3 signaling. Nrip1 and Rex3 were proposed to be diagnostic markers of different forms of leukemias 
. Vilar et. al. suggested that Rex3 links cell surface receptor signaling to the cell cycle and neuronal differentiation by binding the neurotrophin receptor 
; HSC-selective expression indicates that Rex3 could play similar roles in regulating proliferation and differentiation among hematopoietic precursors. Smarca2, a SWI/SNF-related regulator of chromatin also known as Brm, plays important roles in transcriptional processes via regulation of chromatin modifications and structure. Smarca2 also interacts with TopBP1, a protein involved in DNA replication and the DNA damage checkpoint to promote cell survival 
. Smarca2 has been linked to inhibition of cell proliferation through interactions with cyclin D3 
. In agreement with the differential expression in our data sets, Muchardt et al suggested previously that increased levels of Smarca2 promote the withdrawal of the cell from the cell cycle 
Collectively, the genes discussed above define a molecular signature associated with the steady-state quiescent status of HSC, and suggest common targets whose perturbation can lead to enhanced cell proliferation in the context of either differentiation, stem cell expansion, or transformation.
Transcriptional Regulators of HSC-Specific Genes
To identify potential master regulators of genes selectively expressed in normal HSC, we performed sequence analysis of the genomic regions upstream of the 93 genes in the common intersect. We searched for both known transcription factor binding sites using the Transfac database (http://www.cbil.upenn.edu/cgi-bin/tess/tess?RQ=NBqt
) and for enrichment of novel sequence motifs. The abundance of these sequences in the 3way intersection was compared to both the genome-wide input dataset and to the mouse genome. This analysis identified two sequence motifs enriched with statistical significance in the HSC-selected upstream regions (). Intriguingly, these sequences may be bound by as of yet unidentified transcriptional regulators, as they do not contain sequence motifs recognized as binding sites for known transcription factors. In addition, we identified binding sites for 467 known transcription factors and sites for 8 of these were significantly enriched in the HSC dataset (). Literature searches revealed that all 8 transcription factors have been implicated in hematopoiesis. In particular, Evi1 is an oncogenic transcription factor in myeloid leukemias, and may regulate normal hematopoiesis by interacting with transcription factors in the Gata family 
. Deletion of the ubiquitously expressed basic leucine zipper transcription factor AFT4 leads to severe anemia 
. The interferon response gene IRF1 may act as a tumor suppressor by regulating the differentiation and proliferation of myeloid cells via the Stat pathway 
. Nf1 is a tumor suppressor that regulates normal hematopoiesis. Mutations in or deletion of Nf1 leads to myeloproliferative disorders 
. NF-Y may play roles in the control of HSC self-renewal by inducing the expression of Hoxb4 
, a known positive regulator of HSC self-renewal 
as well as other genes involved in HSC self-renewal 
. NF-Y also has been shown to regulate histone methylation and may function as both an activator and repressor of transcription 
. Finally, the Ikaros proteins are well characterized in hematopoiesis 
. Altered ik2, ik4 and ik6 transcripts have recently been identified in Philadelphia chromosome-positive acute lymphoblastic leukemia (Iacobucci 2008), demonstrating that Ikaros and Ikaros target genes play important roles in leukemogenesis.
Putative transcription factor binding motifs enriched in upstream regulatory regions of genes selectively expressed by normal, quiescent HSC.
Transcription factors with binding sites enriched in the upstream region of genes expressed selectively in HSC.
It is intriguing to note that the transcription factors we identified in this metaanalysis play important roles in hematopoiesis. It will be of interest to determine the comprehensive set of target genes for these transcription factors in HSC to further build a molecular portrait of the complex networks of genes that regulate normal HSC function.