Transcriptional regulation in myelopoiesis: Hematopoietic fate choice, myeloid differentiation, and leukemogenesis

Transcription

1 Experimental Hematology 33 (2005) Transcriptional regulation in myelopoiesis: Hematopoietic fate choice, myeloid differentiation, and leukemogenesis Alan G. Rosmarin a,b, Zhongfa Yang b, and Karen K. Resendes b a Department of Medicine, Brown Medical School, Providence, RI, USA; b Department of Molecular Biology, Cell Biology, and Biochemistry, Brown University, Providence, RI, USA (Received 6 August 2004; accepted 6 August 2004) Myeloid cells (granulocytes and monocytes) are derived from multipotent hematopoietic stem cells. Gene transcription plays a critical role in hematopoietic differentiation. However, there is no single transcription factor that is expressed exclusively by myeloid cells and that, alone, acts as a master regulator of myeloid fate choice. Rather, myeloid gene expression is controlled by the combinatorial effects of several key transcription factors. Hematopoiesis has traditionally been viewed as linear and hierarchical, but there is increasing evidence of plasticity during blood cell development. Transcription factors strongly influence cellular lineage during hematopoiesis and expression of some transcription factors can alter the fate of developing hematopoietic progenitor cells. PU.1 and CCAAT/enhancer-binding protein a (C/EBPa) regulate expression of numerous myeloid genes, and gene disruption studies have shown that they play essential, nonredundant roles in myeloid cell development. They function in cooperation with other transcription factors, co-activators, and co-repressors to regulate genes in the context of chromatin. Because of their essential roles in regulating myeloid genes and in myeloid cell development, it has been hypothesized that abnormal expression of PU.1 and C/ EBPa would contribute to aberrant myeloid differentiation, i.e. acute leukemia. Such a direct link has been elusive until recently. However, there is now persuasive evidence that mutations in both PU.1 and C/EBPa contribute directly to development of acute myelogenous leukemia. Thus, normal myeloid development and acute leukemia are now understood to represent opposite sides of the same hematopoietic coin International Society for Experimental Hematology. Published by Elsevier Inc. Granulocytes and monocytes are key mediators of innate immunity and the inflammatory response. Collectively, these cells and their committed progenitors are referred to as myeloid cells. The term myeloid is derived from the Greek word for marrow and, indeed, myeloid cells constitute the dominant cellular population in bone marrow. Myeloid cells arise from multipotential hematopoietic stem cells (HSCs), which constitute a rare population in the bone marrow estimated to be approximately one per 10 5 marrow cells [1]. As with all stem cells, HSCs have the ability to differentiate into multiple cell types and to replenish the pluripotent population ( self-renew ). The differentiation of HSCs to various hematopoietic lineages has been studied intensively and the mechanisms that regulate these processes provide important Offprint requests to: Alan G. Rosmarin, M.D., The Miriam Hospital, 164 Summit Avenue, Providence, RI 02906; Rosmarin@Brown.edu models for the regulation of fate determination in other types of stem cells. Myeloid cells are replete with proteins that mediate recognition, ingestion, and destruction of foreign organisms, antigen presentation, cytokine production, and other functions of the immune and inflammatory reactions. Most genes that are regulated during myeloid differentiation are controlled at the transcriptional level. Thus, identification of regulatory DNA sequences and the transcription factors that bind these sites provides important insight into the molecular mechanisms that control myeloid cell development. As will be discussed, there is no single, master regulator of transcription in myeloid cells. Rather, transcriptional regulation in myeloid cells is the result of the combinatorial effects of a few key transcription factors [2]. This review will not attempt to serve as an all-inclusive summary of myeloid genes and their transcriptional regulation; other excellent reviews of myeloid cell development are X/05 $ see front matter. Copyright 2005 International Society for Experimental Hematology. Published by Elsevier Inc. doi: /j.exphem

2 132 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) Figure 1. Conventional model of hematopoietic differentiation. hematopoietic stem cell (HSC), common lymphoid progenitor (CLP), common myeloid progenitor (CMP), and mature elements of blood are indicated. available [3 6]. It will, however, emphasize the transcription factors, PU.1 and CCAAT/enhancer-binding protein α (C/EBPα), because of their crucial roles in controlling myeloid gene expression and cellular development. The increasing recognition of hematopoietic plasticity and the role of transcription factors in this phenomenon will also be addressed. Finally, we will review the convergence of studies of normal myelopoeisis and cancer biology, which demonstrate that PU.1 and C/EBPα regulate normal myeloid development, while their aberrant expression contributes to pathogenesis of acute myelogenous leukemia (AML). Stem cell commitment to myeloid fate A classical understanding of the development of mature blood cell types from HSCs is presented in Figure 1. This model incorporates findings from immunophenotyping, in vitro culture analyses, experimental bone manrow transplantation, and clinical experience. In this view, hematopoiesis is a relatively linear and hierarchical process whereby pluripotent HSCs undergo successive symmetric and asymmetric divisions to yield committed progenitor cells. The latter also possess stem cell-like properties, but exhibit a progressive restriction of cellular fate as they differentiate toward mature lymphoid, erythroid, megakaryocytic, or myeloid cells. Committed progenitor cells proliferate to meet the enormous daily needs of blood cell production and through this hierarchical system, they ultimately yield the mature elements of blood [7,8]. Abundant evidence supports this hierarchical view of hematopoietic differentiation. For example, it is possible to isolate populations of highly purified progenitor cells that demonstrate a high degree of lineage fidelity in vitro and in vivo, such as common lymphoid progenitor cells (CLPs)[9] and common myeloid progenitor cells (CMPs) [10]. For decades, HSCs were more of a theoretical concept than a physical entity that could be identified and manipulated. Unlike their mature circulating progeny lymphoid, erythroid, megakaryocytic, or myeloid cells HSCs cannot be definitively identified by morphologic criteria. In fact, HSCs appear as relatively undistinguished mononuclear cells. However, in recent years the ability to prospectively identify, isolate, and characterize HSCs and committed progenitor cells by immunophenotyping and metabolic properties has provided important tools for the study of hematopoietic differentiation. Mature blood cells express characteristic cell surface markers, i.e., CD11b for myeloid cells, CD3 for T lymphocytes, B220 for B cells, and Ter- 119 for erythroid cells. HSCs do not express substantial amounts of these lineage markers, and this property has been exploited to enrich bone marrow for HSCs. Cells that express such differentiation markers can be removed, and the remaining, lineage-negative cells (Lin ) are enriched up to 100-fold for HSCs. Conversely, less-differentiated cells express the antigens CD34, c-kit, and Sca-1, and this property can be used to positively select cell populations that contain

3 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) HSCs. Metabolic characteristics, such as low uptake of rhodamine dye or Hoechst stain can also be used to enrich HSCs. Typically, some combination of these approaches is used to isolate immature bone marrow cells. These strategies permit prospective isolation of subsets of progenitor cells for analysis of growth characteristics, gene expression, and regulators of cell fate, survival, and differentiation. The figure indicates several key branchpoints in hematopoietic cellular fate. Individual cells that are committed to either the myeloid or lymphoid fate, as judged by in vitro culture assays and experimental bone marrow transplantation, can be isolated by immunophenotype. For example, CLPs can be isolated based on expression of interleukin (IL)-7 receptor [9] and their progeny can be further distinguished for B and T cell fates [11]. Similarly, CMPs can be distinguished by immunophenotype, i.e., expression of CD34 and FcγR can be used to isolate populations that are committed to a granulocyte-monocyte (GM) fate vs a megakaryocytic-erythroid (Meg-E) fate [10]. Thus, the ability to prospectively isolate specific populations of progenitor cells provides powerful tools for the analysis of the transcription factors and cellular events that drive hematopoietic differentiation. This model implies a fixed and definable pattern of gene expression and chromatin configuration that, once selected, is relatively immutable. However, such a static representation of hematopoiesis is incomplete, for it is increasingly evident that fate commitment is highly dynamic. This plasticity is most evident in the emerging body of literature that describes the ability of HSCs to assume nonhematopoietic fates. Various animal models have demonstrated that HSCs can repopulate nonhematopoietic cellular compartments, including liver, muscle, and other tissues. The mechanisms of repopulation and cellular plasticity remain highly contentious, as competing models attempt to distinguish between transdifferentiation, i.e., conversion to an altogether different lineage fate, and cell fusion, in which bone marrow-derived cells fuse to repopulate a target cell population [12,13]. Fate commitment within the hematopoietic compartment is also recognized to exhibit plasticity and the mechanisms that account for fate conversion within the hematopoietic compartment reflect the larger debate. What are the factors that influence fate choice in hematopoiesis? Decades of experimental work have shown that lineage commitment is strongly influenced by cytokines. Agents such as granulocyte-colony stimulating factor (G-CSF), monocyte-csf (M-CSF), and GM-CSF foster differentiation toward their namesake lineages in vitro and in vivo. The identification of cytokines that support lineage development has been a scientific and therapeutic triumph, and several of these compounds are in routine clinical use. Hematopoiesis is also influenced by interactions between developing blood cells and bone marrow stroma, but these molecular interactions are less well defined. The role of cytokines in directing individual cells to particular hematopoietic fates has been intensively debated for decades. One view suggests that lineage commitment is stochastic i.e., a random process. In this view, cytokines provide survival signals that rescue cells from alternative lineage fates, death, or apoptosis [14]. In contrast, instructive or deterministic models posit that cells are driven toward a particular fate by growth factors, interactions with stroma, or other external influences, [15]. Alternative models have also been proposed. For example, hematopoiesis has been envisioned as a series of successive binary choices [16]. Others have suggested more nuanced fate choices, and envision hematopoiesis as a fluctuating continuum of stem and progenitor cell potential that varies with the cell cycle and is influenced by cell surface antigen expression and environmental context [17,18]. The stochastic vs instructive views are typically proposed as competing, dichotomous models. The ongoing debate over these competing models persists because of powerful experimental data to support aspects of each of these models and, ironically, the same data are often invoked to support the various models. However, studies of hematopoiesis in transgenic mice and in vitro manipulations of progenitor cells have eroded the clear lines of these classic models and they may be evolving into models that incorporate aspects of each approach. Neatly packaged schemata of hematopoiesis with sharply defined lineage fate choices written in black or white are yielding to more dynamic models. Chiaroscuro, which evokes the subtly shaded painterly effects of Renaissance art, may be a more apt metaphor than the cleanly drawn cartoons by which hematopoietic hierarchy is currently defined [17]. Regardless of whether one favors a stochastic, deterministic, or other model, it is clear that transcription factors play key roles in determining the ultimate fate of a differentiating hematopoietic cell. Transcription factors drive the patterns of gene expression that are characteristic of a given hematopoietic lineage, i.e., the myeloid genes that mediate the inflammatory and immune responses of these cells. One might envision HSCs as blank slates into which a lineage-specific program will be carved. Alternatively, they can be viewed as a scratch pad on which an array of transcriptional programs is drafted, but not yet finalized. The former model would point toward a linear correspondence of lineage and specific markers, while the latter, i.e., lineage priming would suggest more promiscuous patterns of gene expression. In either view, transcription factors are critical mediators of differentiation. Gene expression in HSCs If hematopoiesis is linear and hierarchical, then gene expression should proceed in an orderly pattern. Thus, as HSCs differentiate into committed progenitors and mature blood cells, there should be an orchestrated activation of some genes and

4 134 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) simultaneous silencing of others. HSCs presumably express genes that encode the characteristic properties of stem cells, i.e., the ability to self-renew and differentiate, home to bone marrow, etc. Several approaches have been used to identify genes that are preferentially expressed by stem cells, including insertional mutagenesis [19], differential display [20], representation difference analysis (RDA) [21], and microarrays [22]. Presumably, some of these preferentially expressed genes [23] encode the characteristic phenotypic properties of HSCs. As HSCs differentiate, these stem cell genes should be gradually silenced and replaced by genes that reflect the selected lineage, i.e., characteristic myeloid or erythroid genes. However, such orderly patterns of gene expression are not the rule in hematopoiesis. Analysis of gene expression in the multipotent hematopoietic progenitor cell line, FDCP mix, and in human CD34 Lin cells indicates that single cells co-express characteristic genes from divergent lineages, i.e., simultaneously express erythroid genes such as globins, and myeloid genes, such as myeloperoxidase (MPO) [24]. Furthermore, cytokine receptor genes, which are ostensibly lineage-specific, were also expressed by a broader than expected range of developing hematopoietic cells [25]. Similar findings were observed with freshly isolated HSCs and lineage-committed progenitors [26]. Ye and colleagues [27] recently used an alternative approach to further reinforce this observation. They used a Cre-loxP based approach to elegantly demonstrate that the myeloid gene, lysozyme, is expressed not only in myeloid cells, as expected, but also in lymphoid cells and in repopulating HSCs [27]. Thus, reverse transcription-polymerase chain reaction (RT-PCR) and microarray studies indicate that HSCs and their progeny simultaneously express unexpected combinations of lineage-restricted genes. Interestingly, this promiscuous gene expression in developing cells is not random. Individual cells express unexpected combinations of ostensibly lineagerestricted hematopoietic genes, e.g., globins and MPO, transcription factors, and growth factor receptors but, importantly, they do not express patterns of genes from unrelated programs such as myogenic profiles [24]. This argues against general leakiness of gene expression. Rather, it is as if the cells were having a hematopoietic identity crisis, rather than exhibiting a global state of confusion. In reality, most so-called lineage-restricted genes may be expressed more widely than was previously appreciated. However, HSCs appear to express these lineage-specific markers at very low levels, compared with expression by more differentiated cells. Expression is presumably below the threshold of detectability by cell sorting (which explains why Lin populations of bone marrow cells are enriched for HSCs), yet can be detected by more sensitive techniques, such as RT-PCR. Alternatively, these genes may be transcribed, but not efficiently translated and expressed. Thus, low levels of expression of lineage-associated genes by HSCs may provide a sneak preview of factors that ultimately govern fate choice. Low-level expression of such lineage-restricted products may undergo random fluctuation to cause a self-reinforcing pattern of gene expression. Several key transcription factors, i.e., GATA-1 and PU.1, regulate their own expression [28,29]; such autoregulation may reinforce early forays along a differentiation lineage. These fluctuating and then self-reinforcing patterns of gene expression may provide a mechanistic explanation for the apparently stochastic behavior of HSCs. Lineage commitment and fate switching What constitutes lineage commitment? Is it an absolute step, or does it represent a dynamic state that is at least partially reversible? It is clear that biochemical manipulations and altered cytokine activity can redirect cells to a myeloid lineage fate. Treatment of pre-b lymphoma cell lines with azacytidine-induced properties of macrophages, including adherence to plastic, phagocytosis, and esterase activity [30]. Expression of v-raf in Eµ-myc transformed pre-b cells had similar functional and gene expression effects [31]. Similarly, expression of v-fms (a constitutively activated form of M-CSF receptor) switched B cell lines to macrophages [32]. Such lineage switching is not restricted to B lymphoid cells, for enforced expression of GM-CSF or IL-2 can drive CLPs and pro-t cells towards a GM fate [33,34]. Transcription factors also play critical roles in directing lineage commitment. In nonhematopoietic systems, this may be best illustrated in studies where expression of MyoD was sufficient to generate muscle cells from a range of cell types, including fibroblasts and retinal epithelial cells [35,36]. The most striking effect of a transcription factor in defining hematopoietic lineage is the role of Pax5 in early development of the B cell lineage [37]. Pax5 not only increases expression of the characteristic B cell genes, CD19 and Igα, but it also suppresses myeloid genes, such as c-fms and PD-1. In the absence of Pax5, B cell precursors exhibit greater lineage plasticity, and can develop into T lymphocytes, natural killer cells, and macrophages [38]. C/EBPα and GATA transcription factors can switch Pax5 null cells, but not wild-type cells, from a lymphoid to a myeloid fate [39]. Thus, Pax5 both activates the B cell program and represses other lineage fates. Forced expression of crucial transcription factors can also redirect myeloid-lineage fate. For example, expression of GATA-1 in Myb-Ets transformed hematopoietic progenitor cells reprograms them into eosinophils or thromboblasts, while at the same time, it suppresses myelomonocytic differentiation [40]. Conversely, induction of PU.1 in these cells activates myeloid gene expression and causes irreversible myeloid differentiation while downregulating GATA-1 expression and suppressing alternate lineage fates [41]. Similarly, GATA-1 expression in HSCs induces Meg-E differentiation and reprograms CLPs and GM progenitors to

5 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) differentiate into Meg-E lineages, while inhibiting normal lymphoid or GM differentiation [40,42]. Within the myeloid lineage, forced expression of C/EBPα in the bipotential U937 myeloid cell line triggers granulocytic differentiation while suppressing the monocytic differentiation program [43]. Forced expression of C/EBPα or C/EBPβ in B lymphocytes reprogrammed them to a macrophage fate [44]. These studies point to substantial plasticity of hematopoietic cells. They demonstrate that fate commitment is both a positive action toward a particular fate and, equally importantly, active repression of alternate fate choices. Furthermore, they strongly support the concept that transcription factors play a deterministic role in hematopoietic fate choice. Chromatin and lineage restriction The final substrate for DNA-based processes in vivo is not just DNA sequence, it is chromatin. Lineage-specific gene expression requires more than simply providing the necessary transcription factors; critical regulatory elements must be accessible in the context of chromatin. Remodeling of chromatin occurs as cells differentiate, until an irreversible state is achieved that is specific for a terminally differentiated cell. Rapidly proliferating cells, such as hematopoietic progenitor cells, face the additional challenge of maintaining the proper chromatin context as they replicate their DNA during passage through the cell cycle. Thus, maintenance of proper chromatin context may be as important as the availability of transcription factors for determining lineage fate. Chromatin is composed of DNA that is tightly wrapped around histone proteins, which comprise the structural core of nucleosomes. Several epigenetic modifications are known to alter chromatin structure. The tail regions of histones are subject to acetylation, methylation, phosphorylation, ubiquitination, and adenosine diphosphate-ribosylation. DNA, itself, is subject to modification in the form of methylation. In recent years, it has become apparent that a histone code, which is written in the language of post-translational modifications of histones, alters the conformation and functional properties of chromatin [45]. These modifications alter the electrostatic charge of histones and create binding surfaces that recruit functional complexes to their proper sites of action. These epigenetic effects and the recruited protein complexes modulate transcriptional activation and repression, DNA replication, recombination, and repair. These epigenetic effects are influenced by the cell type, state of differentiation, cell cycle status, and other environmental cues. Many of the transcription factors that regulate myeloid genes interact with transcriptional coactivator and represser molecules, which can confer epigenetic changes on chromatin. For example, the transcriptional coactivator p300 binds to several key myeloid transcription factors, including PU.1, C/EBPα, and retinoid receptors [46]. Because p300 binds to different transcription factors through distinct interaction domains, it may serve as a platform for assembly of multiprotein complexes. p300 possesses intrinsic histone acetyl transferase (HAT) activity by which it modifies chromatin [46].It is also capable of directly acetylating transcription factors. For example, p300 acetylates AML1 a key activator of important myeloid genes and this modification has a substantial effect on the DNA-binding activity of AML1 [47]. This is not a one-way street, for interaction with transcription factors can also lead to modifications of p300; e.g., physical interactions with C/EBP factors leads to phosphorylation of p300, which alters its coactivator function [48]. Conversely, corepressor molecules possess histone deacetylase (HDAC) activity, which removes acetyl groups from histones and transcription factors [49]. Yet another layer of epigenetic complexity is the cross-talk between methylation of DNA and histone modifications. DNA methylation is linked to transcriptional repression, including genomic imprinting and transcriptional regulation during development [50]. The nature of many chromatin modifications is now largely catalogued, but there are more subtle effects yet to be decoded. For example, the number of methylated residues per histone molecule may determine if a gene is activated or repressed, and which specific residue is altered can have diametrically opposite effects [51,52]. The net effect of these chromatin modifications is that the accessibility of genes to transcription factors and other regulatory proteins is altered. The low-level transcription of lineage-specific genes found in HSCs and committed progenitors may be due to selective priming ( opening up ) of chromatin in these precursor populations. In myeloid cells, this is best illustrated with the lysozyme gene. In chickens, chromatin reorganization of the lysozyme locus begins before onset of gene expression, and the fine structure chromatin characteristics of lysozyme-expressing macrophages are seen in multipotent myeloid precursor cells [53]. However, transcription factors only transiently interact with this locus in immature cells [54]. As they differentiate to mature macrophages, complete demethylation of the locus and more complete binding of transcription factors is seen [55]. Thus, the absence of full-scale remodeling at these loci (as is seen in mature, fully differentiated cells) appears to permit only a trickle of gene expression. This may account for the low-level expression of key transcription factors that occurs in progenitor cells, and may thus be a key to stochastic models of differentiation. Thus, the role of chromatin modification in myeloid development is increasingly evident. As described earlier, it has been known for decades that the DNA-demethylating agent, azacytidine, can redirect hematopoietic fate from pre-b cells to macrophages [30]. Many of the key myeloid transcription factors regulate gene expression, at least in part, through their interactions with HATs and HDACs. The importance of these interactions is further underscored by alterations of these coactivators and corepressors in leukemia. The HAT coactivator, CBP, is directly involved in

6 136 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) chromosomal rearrangements that generate fusion proteins which are seen in leukemia [56]. Furthermore, several chimeric proteins that are associated with leukemia, including AML-1/ETO and promyelocytic leukemias that generate fusions of retinoic acid receptor (RAR)-α directly interact with corepressors with HDAC activity [57]. Thus, chromatin modification appears to play important roles in fate choice and differentiation of normal myeloid cells, and in the pathophysiologic state of acute myeloid leukemia. PU.1 and C/EBPa: two critical myeloid transcription factors As described earlier, transcription factors play a pivotal role during myeloid differentiation. However, there is no single master myeloid transcription factor that alone governs myeloid lineage commitment, as seen with MyoD in muscle cells or Pax5 in B lymphocytes. Instead, multiple transcription factors work cooperatively and coordinately to regulate both temporal and lineage-specific gene expression [2]. More than a dozen transcription factors play important roles in regulating myeloid genes [3 6]. This review will focus on PU.1 and C/EBPα because they regulate numerous myeloid genes and they are specifically required for development of the myeloid lineages. Furthermore, recent findings in patients with AML have confirmed a long-held belief that mutations of transcription factors that are required for normal myeloid differentiation should be associated with the development of acute leukemia. Why, then, are PU.1 and C/EBPα not myeloid master transcription factors? This definition should be reserved for a factor that is restricted to a specific cellular lineage and that, alone, is sufficient to control gene expression and direct differentiation fate. Neither PU.1 nor C/EBPα is truly myeloid restricted, i.e., PU.1 is also expressed by B lymphocytes and C/EBPα is expressed by liver and other organs. Thus, although myeloid gene expression is achieved by synergistic regulation by multiple transcription factors, both PU.1 and C/EBPα play critical roles in myeloid differentiation and they deserve special attention. PU.1 PU.1 is a member of the ets transcription factor family. The term, ets, is an acronym that recalls the founding member of this family Ets1, which was first identified from the E26 (E twenty six), hence Ets, retrovirus [58,59]. The namesake ets region refers to the characteristic 85 residue wingedhelix-turn-helix domains that bind to DNA sequences with a GGA core [60 62]. There are more than 20 distinct mammalian ets factors, and they differ considerably in their transcriptional activation domains and other functional regions [63 65]. PU.1 is among the more distantly related ets factors it shares only 35% amino acid identity with Ets1. PU.1 also includes a PEST domain, which is associated with protein instability [66]. PU.1 was first described in Friend virusinduced erythroleukemia, and its overexpression blocks erythroid differentiation [67]. In cell lines, PU.1 is expressed by myeloid cells and B lymphocytes, but not by T cell lines [68,69]. Its expression increases during differentiation of granulocytic and monocytic precursors [70]. In murine bone marrow, it is expressed by myeloid and erythroid precursors, macrophages, and megakaryocytes, but it is not found in mature granulocytes, osteocytes, or vascular endothelium [71]. Binding sites for PU.1 are found on almost all myeloid specific promoters. Notable target genes are the characteristic antigens, CD11b and CD18 [72,73] and the receptors for the cytokines, M-CSF, GM-CSF, and G-CSF [74 77]. In addition, PU.1 binds and regulates its own promoter [29]. Recognition that PU.1 is expressed by myeloid cells and that it binds to numerous myeloid promoters suggested an important role in myeloid differentiation. Its importance in myeloid cells was confirmed when PU.1 was disrupted in mice by two different groups. Their experimental strategies differed, and they observed somewhat different results. However, both groups found that loss of PU.1 disrupted development of both myeloid cells and B lymphocytes; they differed in the observed effects on T cell development [78,79]. Singh s group generated PU.1 deficient mice by replacing most of the ets DNA binding domain with the neomycinresistance gene [78,80] PU.1 disruption caused embryonic lethality between days 16.5 and 18.5 of gestation. Morphological analysis and flow cytometry revealed no detectable myelocytes or lymphocytes. Clonogenic assays on mutant myeloid progenitors also confirmed the loss of myeloid progenitors in PU.1 ( / ) fetal liver. These studies indicated an early differentiation block on both myelopoiesis and B lymphopoiesis. Meanwhile, Maki s group disrupted PU.1 by inserting the neomycin-resistance gene into the ets domain [79]. PU.1 targeted pups were born with the expected Mendelian ratio, but they suffered septicemia within 2 days; mutant pups could survive up to 2 weeks with antibiotic therapy. Flow cytometry on bone marrow and fetal liver cells of those neonates detected no mature granulocytes or macrophages. A small number of aberrant neutrophils and macrophages was found in the bone marrow and spleen of the older pups, indicating a partial, yet severe impairment in myelopoiesis. These mice also exhibited a complete block in B cell differentiation and a delay in T cell maturation. Thus, both PU.1 null mouse models showed severe abnormalities in myeloid cell development. PU.1 null embryonic stem cells fail to express a variety of characteristic myeloid genes, including CD11b and CD18 [81]. Hematopoietic cells derived from PU.1 null neonates and fetal liver can generate neutrophil colonies [82,83]. They generate cells that exhibit some characteristics of monocytes, such as phagocytosis and expression of characteristic antigens, but they failed to develop into mature monocytes [84].

7 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) PU.1 null cells do not express the receptors for M-CSF, GM- CSF, and G-CSF [82,83,85]. Reintroduction of PU.1 into PU.1 null hematopoietic progenitors cells by retrovirusmediated expression rescued myeloid and lymphoid development in vitro [86] but introduction of G-CSF receptor (G-CSF-R) or M-CSF-R was not sufficient to rescue granulocyte or monocyte development [82,83]. PU.1 expression increases during myeloid differentiation and it may be required for terminal maturation of myeloid cells [70]. Treatment of monocytes with GM-CSF induces macrophage differentiation and causes accumulation of PU.1. Interestingly, transduction of alveolar monocytes with a PU.1-expressing retrovirus was sufficient to drive macrophage differentiation in the absence of GM-CSF [87]. A similar effect is seen with maturation of granulocytes. Expression of PU.1 increases as immature myeloid cells differentiate into mature granulocytes [88]. In 32D murine myeloid cells, G-CSF drives increased PU.1 expression via a STAT-dependent mechanism, and expression of dominant negative STAT both blocks PU.1 upregulation and granulocytic differentiation [89]. Thus, studies of PU.1 expression and its genetic disruption suggest that hematopoietic cells can commit to the myeloid lineages in the absence of PU.1, but they fail to mature in a normal manner. Physical and functional interactions of PU.1 with partner proteins are critical to its role in gene expression and lineage determination. c-jun cooperates with PU.1 in monocytic differentiation, either by binding to adjacent DNA elements or by direct physical interactions. Expression of c-jun, alone, induced partial monocytic differentiation of a variety of myeloid cell lines [90 92]. PU.1 and c-jun physically interact via their DNA binding domains and cooperate to activate gene transcription [93]. In contrast, PU.1 and the erythroid transcription factor, GATA-1, mutually antagonize their respective transcriptional activities. PU.1 can bind to the DNA binding domain of GATA-1 and block its ability to bind to DNA. Conversely, GATA-1 binds to PU.1 and displaces c-jun, thereby reducing expression of target genes [94 97]. Because PU.1 and GATA-1 autoregulate their own expression, modest fluctuations in transcription factor expression may translate into significant functional effects in lineage commitment and gene expression. For example, expression of PU.1 in erythroleukemia cells downregulated GATA-1 expression and caused a switch in lineage fate toward myeloid differentiation [98]. These results indicate that PU.1 and GATA-1 not only activate their distinct lineage specific target genes, but also repress the target genes of the other lineage via protein protein interactions. The physical and functional interactions of PU.1 in B lymphocytes are strikingly different. In B cells, PU.1 interacts with interferon response factor-4 (also known as Pip), which is an essential cofactor in B lymphocytes [99]. In addition, Singh s group made the important observation that levels of PU.1 expression may influence fate choice between myeloid cells and B cells. They reintroduced PU.1 along with green fluorescent protein (GFP) into PU.1 null cells and rescued development of both lineages. Interestingly, they found that CD11b macrophages expressed high levels of GFP, whereas pro-b cells expressed low levels of GFP. They suggested that high levels of PU.1 induce commitment to the myeloid lineage, while lower levels drive B lymphoid development [86]. Studies of the multipotent, factor-dependent hematopoietic cell line, FDCP-mix, support this concept. Transient expression of PU.1 in FDCP-mix cells drove development of both granulocytes and macrophages and blocked erythroid development. Interestingly, stable expression of PU.1 fostered macrophage development and blocked granulocytic differentiation [100]. This suggested that sustained, high-level expression of PU.1 drives myeloid differentiation and favors macrophage development over granulocytic development. As will be described here, levels of PU.1 expression may also be critical to normal vs leukemic development in myeloid cells. C/EBPa C/EBPα is the founding member of a family of related leucine-zipper transcription factors that play important roles in myeloid differentiation. These proteins share related N-terminal transactivation domains, basic DNA binding regions, and C-terminal leucine-zipper protein interaction domains. C/EBP factors homo- and heterodimerize via their leucine zipper domains and bind to DNA via their adjacent basic regions [101]. There are six members of the C/EBP family: α,β,δ,ε,γ, and CHOP. Further complexity arises from the expression of multiple isoforms that are created by alternative splicing and alternative translational initiation sites. Because some of these proteins and isoforms exclude the transcription activation domains, heterodimerization can yield dizzying numbers of complexes with different functional capacities [ ]. None of the C/EBP factors is restricted to myeloid cells, for they are also expressed in other hematopoietic lineages, liver, adipocytes, and other tissues. High-level expression of C/EBPα, β, and δ is found in granulocytes, monocytes, and eosinophils [43,106,107]. C/EBPα is the predominant isoform in immature granulocytes [76,106]. In contrast, C/EBPε is found predominantly in maturing granulocytes and T lymphocytes [108]. CHOP is expressed only in granulocytes that are subjected to stress, such as DNA damage [109]. There are multiple, functionally important targets of the C/EBP factors in myeloid cells, including G-CSF-R, M- CSF-R, and GM-CSF-R [76,77,110]. C/EBPs also regulate important components of the inflammatory response such as CD14; primary granule protein genes, such as myeloperoxidase, lysozyme, and neutrophil elastase; and secondary granule protein genes, such as lactoferrin and neutrophil collagenase [ ]. This review will focus on C/EBPα because of its important and nonredundant roles in myeloid cell development, function, and pathophysiology. C/EBPα is abundant in early

8 138 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) myeloid cells where it binds and activates key myeloid target genes. The expression of C/EBPα mrna and protein by early myeloid cells increases up to threefold following induction of granulocytic differentiation by retinoic acid in myeloid cell lines; in contrast, it is rapidly downregulated during monocytic differentiation [43]. These changes in expression are also seen in normal human granulocytes [43] and in analysis of single, primary human hematopoietic cells [70]. Conditional expression of C/EBPα in the factor-dependent granulocytic cell line, 32Dcl3, induced granulocytic differentiation. These cells expressed both early granulocytic genes such as G-CSF-R, MPO, and lysozyme, and late granulocytic genes, such as lactoferrin and C/EBPε [117]. Conditional expression of C/EBPα in the biopotential myeloid cell line, U937, also led to granulocytic morphology and expression of mrna for G-CSF-R, lactoferrin, and neutrophil collagenase; it also blocked monocytic differentiation. Interestingly, induction of C/EBPα for as little as 48 hours was sufficient to induce these changes, but the cells required more than 2 weeks to achieve these effects [43], presumably due to the initiation of a cascade of secondary events. Introduction of C/EBPα into avian multipotential progenitor cells also upregulated myeloid markers [118]. Thus, upregulation of C/EBPα is sufficient to activate characteristic granulocytic morphology and gene expression. It should be noted, however, that other C/EBP factors also induce granulocytic differentiation in myeloblastic cell lines [ ]. C/EBPα serves a nonredundant role in early granulocyte development. Tenen s group reported that C/EBPα null neonates lacked mature granulocytes, but had normal erythrocytes, megakaryocytes lymphocytes, and monocytes/ macrophages. There was a significant increase of neutrophil precursors in the peripheral blood and a selective loss of colony-forming unit (CFU)-G colonies, but normal CFU- M, CFU-GM, and CFU-GEMM colonies. Northern blots detected no G-CSF-R mrna expression in granulocytes of mutant pups, and treatment with G-CSF did not rescue neutrophil development [123]. Thus, disruption of C/EBPα in mice resulted in an early and specific differentiation block of granulocytes, indicating its indispensable role in early granulocytic commitment. Some of the target genes that account for this block in granulocytic differentiation by C/EBPα deletion have been identified. Restoration of G-CSF-R by retroviral transduction of fetal liver cells partially rescued the granulocyte defect [124]. IL-6 receptor expression was also decreased in the knockout mice, and treatment with IL-6 and soluble IL-6 receptor also partially restored the defect [124]. C/EBPα also regulates expression of PU.1 [117] and given the role of PU.1 in the maturation of later stages of myeloid differentiation (described previously), this may account for some of its effects on full maturation of granulocytes. Other approaches, including RDA, have been used to identify C/EBPα target genes [85]. However, even in the absence of C/EBPα, granulocytic differentiation can be restored by expression of IL-3 and GM-CSF, indicating that there is more than one pathway to maturation of granulocytes. [125,126] The roles of other C/EBP factors in myeloid cells are less well defined. Their genetic disruption also affects hematopoietic differentiation, albeit less dramatically than C/EBPα knockout. C/EBPβ null mice retain all hematopoietic lineages, and the effects of disruption of C/EBPδ, or both C/EBPβ and C/EBPδ are not severe [ ]. Disruption of C/EBPε null mice have immature granulocytes that fail to develop secondary granules [130,131]. Myeloid development and leukemia Many of the same transcription factors that regulate normal myeloid differentiation are now known to participate in the development of leukemia. This was first recognized with RARs. Retinoic acid has long been known to play a role in the differentiation of normal myeloid cells and myeloid cell lines [132]. Dominant negative inhibition of RARα blocks granulocytic differentiation [133]. Rearrangements of RARα are associated with the t(15;17) chromosomal translocation that is seen in acute promyelocytic leukemia [134]. Furthermore, treatment with a retinoic acid derivative, all trans retinoic acid, can induce transient remissions in acute promyelocytic leukemia [135]. Thus, RARs are involved in both normal and malignant (leukemic) myeloid development. Similarly, the AML1 transcription factor is associated with the t(8;21) chromosomal translocation that is commonly found in AML; in this setting it is fused to the ETO gene [136,137]. Interestingly, the natural partner protein of AML1 core binding factor-β, is also involved with a distinct subtype of AML that is associated with chromosome 16 inversions [138]. AML1 is required for expression of early myeloid genes, such as GM-CSF-R, M-CSF-R, MPO, and neutrophil elastase [110, ]. Disruption of either AML1 or CBFβ causes a failure of definitive hematopoiesis in fetal liver with similar phenotypes [ ]. The reason for the different leukemia phenotypes caused by these disruption of these partner proteins is not clear at present. However, the AML1- ETO fusion protein acts at least in part by disrupting function of C/EBPα and PU.1 [147,148]. Given the critical roles of PU.1 and C/EBPα in myeloid differentiation, it was reasonable to expect that abnormalities in these factors, too, would be associated with AML. Until recently, a direct link between leukemia and the transcription factors PU.1 and C/EBPα was elusive. However, recent studies now strongly support this provocative hypothesis. Several groups have now reported C/EBPα mutations with a frequency of 7% to 11% of all AML patients [ ]. Two types of mutations in C/EBPα have been described. The first type involves carboxy terminal mutations that disrupt the basic zipper region, which mediates DNA binding and homo-and hetero-dimerization; some of these cases are associated with a second mutation in the other C/EBPα allele. A second category of C/EBPα mutations involve amino terminal frame shifts that generate prematurely terminated

9 A.G. Rosmarin et al./ Experimental Hematology 33 (2005) C/EBPα; these molecules appear to act in a dominant negative fashion [149]. Thus, point mutations of C/EBPα are found in a substantial proportion of AML patients. Recently, PU.1 mutations were detected in 7% of 126 patients with AML, most of whom had either an undifferentiated phenotype, or myelomonocytic, monocytic, or erythroleukemic differentiation [153]. Because these mutations were generally heterozygous, it is unclear why these individuals developed leukemia, although a gene dose phenomenon has been proposed. Important corroboration for this concept was recently provided by an animal model. Tenen and colleagues created a hypomorphic allele of PU.1 by genetically disrupting a key upstream element that regulates PU.1 expression homozygous mice expressed PU.1 at only 20% of the wild-type levels. These mice expressed poorly differentiated myeloid progenitors and after a relatively short latency of 3 to 8 months, they developed an aggressive disease that resembled AML, which was transferable to immunodeficient and wild-type mice. The additional genetic abnormalities that transform the apparent preleukemic state to overt AML are not yet well defined, although amplification of c-myc was found in four of seven mice [154]. This suggests that reduced levels of PU.1 caused a partial block in myeloid development, and in the presence of additional mutations yielded acute leukemia in a matter of months. The cooperating oncogenes in human PU.1-associated leukemias are not yet undefined. Thus, numerous subtypes of AML are associated with abnormalities of key transcription factors. Most involve reciprocal chromosomal translocations that generate chimeric proteins, which appear to function as rogue transcription factors. Some of these abnormal chimeric factors appear to act, in part, through the transcription factors PU.1 and C/EBPα. More recently, specific point mutations and other abnormalities of C/EBPα and PU.1, themselves, have been found in AML. Although these transcription factors are strongly implicated in leukemogenesis, these defects alone do not appear to be sufficient to cause overt leukemia. Numerous animal models have been generated in which mutant transcription factors are transgenically or retrovirally expressed in hematopoietic cells. Although many of these models generate an abnormal phenotype, none yet fully recapitulates the clinical AML syndrome with which it is associated. For example, various knock-in models of promyelocytic leukemia protein- RARα or promyelocytic leukemia zinc-finger protein RARα caused a myeloproliferative syndrome rather than an acute promyelocytic leukemia [ ]. Similarly, knock-in of AML1-ETO impaired hematopoiesis rather than generating AML [160]. These findings suggest that additional hits are required to transform a normal hematopoietic precursor to acute leukemia. The findings from the hypomorphic PU.1 mouse support the concept that these mutations alone, are not sufficient to generate AML. Just as there is no single master myeloid transcription factor that regulates all myeloid genes, it appears that a defect in a single transcription factor is not sufficient to cause AML. In summary, key transcription factors have been shown to play crucial roles in regulating normal myeloid differentiation and determining fate choice in hematopoiesis. Transcription factors are equally important in malignant myeloid differentiation, insofar as aberrant expression or function of these same factors contributes greatly to AML. Additional events that contribute to leukemogenesis are still being defined, but it is readily apparent that normal myeloid biology and leukemogenesis represent opposite sides of the same coin. Acknowledgments Supported in part by Brown University Herbert W. Savit 49 Research Fund; NIH COBRE 1P20RR017695, NIH COBRE 1P20RR018757, and NIH COBRE 1P20RR (all to A.G.R.). References 1. Metcalf D, Nicola NA. The Hemopoietic Colony-Stimulating Factors: From Biology to Clinical Applicatons. Cambridge: Cambridge University Press; p Shivdasani RA, Orkin SH. The transcriptional control of hematopoiesis. Blood. 1996;87: Zhu J, Emerson SG. Hematopoietic cytokines, transcription factors and lineage commitment. Oncogene. 2002;21: Friedman AD. Transcriptional regulation of granulocyte and monocyte development. Oncogene. 2002;21: Skalnik DG. Transcriptional mechanisms regulating myeloid-specific genes. Gene. 2002;284: Tenen DG. Disruption of differentiation in human cancer: AML shows the way. Nat Rev Cancer. 2003;3: Fuchs E, Segre JA. Stem cells: a new lease on life. Cell. 2000;100: Weissman IL. Stem cells: units of development, units of regeneration, and units in evolution. Cell. 2000;100: Kondo M, Weissman IL, Akashi K. Identification of clonogenic common lymphoid progenitors in mouse bone marrow. Cell. 1997;91: Akashi K, Traver D, Miyamoto T, Weissman IL. A clonogenic common myeloid progenitor that gives rise to all myeloid lineages. Nature. 2000;404: Allman D, Punt JA, Izon DJ, Aster JC, Pear WS. An invitation to T and more: notch signaling in lymphopoiesis. Cell. 2002;109:S1 S Herzog EL, Chai L, Krause DS. Plasticity of marrow-derived stem cells. Blood. 2003;102: Quesenberry PJ, Abedi M, Aliotta J, et al. Stem cell plasticity: an overview. Blood Cells Mol Dis. 2004;32: Enver T, Heyworth CM, Dexter TM. Do stem cells play dice? Blood. 1998;92: Metcalf D. Lineage commitment and maturation in hematopoietic cells: the case for extrinsic regulation. Blood. 1998;92: Brown G, Bunce CM, Lord JM, McConnell FM. The development of cell lineages: a sequential model. Differentiation. 1988;39: Quesenberry PJ, Colvin GA, Lambert JF. The chiaroscuro stem cell: a unified stem cell theory. Blood. 2002;100: Kirkland MA. A phase space model of hemopoiesis and the concept of stem cell renewal. Exp Hematol. 2004;32: Perkins AS, Mercer JA, Jenkins NA, Copeland NG. Patterns of Evi- 1 expression in embryonic and adult tissues suggest that Evi-1 plays