The development of mature granulocytes from hematopoietic precursor cells is controlled by a myriad of transcription factors which regulate the expression of essential genes, including those encoding growth factors and their receptors, enzymes, adhesion molecules, and transcription factors themselves. In particular, C/EBPα, PU.1, CBF, and c-Myb have emerged as critical players during early granulopoiesis. These transcription factors interact with one another as well as other factors to regulate the expression of a variety of genes important in granulocytic lineage commitment. An important goal remains to understand in greater detail how these various factors act in concert with signals emanating from cytokine receptors to influence the various steps of maturation, from the pluripotent hematopoietic stem cell, to a committed myeloid progenitor, to myeloid precursors, and ultimately to mature granulocytes.
Granulopoiesis is a complex process by which primitive blood precursors differentiate into fully differentiated, functionally active granulocytes. Studies on the production of neutrophilic granulocytes and other myeloid cells have provided important paradigms for understanding differentiation. In particular, this work has revealed the intricate and often essential roles played by various transcription factors – both those specific to the granulocytic lineage, as well as more widely expressed molecules – in the control of differentiation. Such transcription factors can act both positively and negatively to regulate the expression of a wide range of important genes, including growth factors and their receptors, other transcription factors, as well as various molecules important for the function of the mature cells. In addition, the activity of several transcription factors is controlled by external stimuli, such as cytokines. Finally, there is a complex interplay between all of these factors – synergistic and antagonistic – which allows for the exquisite control of granulocytic cell production that is observed.
Several genes expressed specifically by immature granulocytic cells have been extensively studied.1234567891011121314 From this work, the transcription factors C/EBPα, PU.1, CBF and c-Myb have emerged as key regulators of gene expression during early granulopoiesis and will be a major focus of this review (Table 1). Furthermore, other studies have implicated several additional transcription factors in the control of granulopoiesis. For example, C/EBPε is expressed predominantly in late-stage granulocytic differentiation and contributes to the expression of a number of key genes,1516 several Hox gene family members are restricted in their expression, within hematopoiesis, to myeloid cells,171819 while the WT-1 gene is expressed in immature hematopoietic cells and in a subset of acute myeloid leukemias.2021 Finally, a number of more ubiquitous transcription factors are known to perform important roles in granulopoiesis in responses to external stimuli: for example, Stat proteins, Myc, and Mad in response to cytokines and the retinoic acid receptor (RAR) in response to retinoic acid.
Transcription factors active during granulopoiesis
C/EBPα (Figure 1) is the founding member of a subset of bZIP transcription factors which include C/EBPβ, C/EBPγ, C/EBPδ, C/EBPε and CHOP.2223242526 C/EBPs form homo- or hetero-dimers via their leucine zipper domains and bind a common DNA element, 5′-T(T/G)NNGNAA(T/G)-3′, via their basic regions.26 C/EBPα is a transcriptional activator27 and has two transactivating domains N-terminal to the bZIP segment.28
CCAAT/enhancer binding protein α (C/EBPα) is expressed in multiple cell types, including adipocytes, hepatocytes and enterocytes.2930 Within the hematopoietic compartment, C/EBPα expression is restricted to the neutrophil, monocyte and eosinophil lineages,313233 being detected in human CD34+CD33+ myeloid progenitor cells, but not in non-myeloid CD34+CD33− cells.33 C/EBPα is the predominant gel-shift species observed in an immature myeloid cell line,7 while its levels decrease during the granulocyte colony-stimulating factor (G-CSF)-mediated granulocytic differentiation of 32D cl3 cells.31 Inducible expression of C/EBPα in the U937 and HL-60 human leukemia cell lines led to the emergence of neutrophilic cells after 2 weeks, with increased expression of the mRNAs for G-CSF receptor (G-CSF-R), lactoferrin, and neutrophil collagenase.33 Similarly, introduction of C/EBPα into multipotent MEP cells led to induction of both myeloid and eosinophil markers.34 C/EBPα linked to the ligand-binding domain of the estradiol receptor (ER), was also introduced into the 32D cl3 cell line.35 When exposed to estradiol, these cells underwent neutrophilic differentiation over a 4-day period, even in the presence of IL-3. MPO RNA induction was detected by 8 h, and G-CSF-R and lactoferrin RNA induction were detected at later times. Consistent with these studies, C/EBPα-null mice lack neutrophils and eosinophils, but retain monocytes and lymphocytes.36 Fetal liver cells from these mice lack detectable G-CSF-R mRNA but retain normal levels of GM-CSF-R and M-CSF-R mRNAs, and hematopoietic colonies are obtained from these cells in response to GM-CSF and M-CSF, but not G-CSF. Colonies grown in GM-CSF contain immature myeloid cells, but not mature neutrophils.36 Expression of early granulocytic or monocytic markers in these cells has not been evaluated. These data indicate that C/EBPα-deficient cells can commit and mature along the monocytic lineage but cannot fully mature along the neutrophilic lineage. In using the phenotypes of C/EBPα-knockout mice to develop a model of the hierarchical regulation of hematopoiesis, one must keep in mind the possibility that other C/EBP family members could compensate for the loss of C/EBPα. Precedent for such compensation exists in GATA-1-knockout mice, in which GATA-2 levels become elevated, enabling further differentiation along the erythroid and megakaryocytic lineages than would perhaps otherwise occur.3738
In contrast to C/EBPα, C/EBPβ levels increase during the differentiation of 32D cells.31 However, C/EBPβ-null mice have reduced B cell numbers and defects in macrophage activation, but otherwise normal myelopoiesis.394041 The ability of C/EBPβ, but not C/EBPα, to cooperate with Sp1 in hepatocytes suggests that a similar collaboration may occur in granulocytic cells.42 C/EBPγ is a ubiquitous protein which lacks a trans-activation domain and can dimerise with other C/EBPs and interfere with their activities.4344 C/EBPε is expressed in T-lineage cells and predominantly in late-stage granulocytes.154546 C/EBPε null mice show defects in neutrophil function, suggesting that C/EBPε controls the expression of a number of key genes.16 This is consistent with a recent study showing that neutrophil-specific granule deficiency results from a loss of function in C/EBPε.47 CHOP contains a defective basic region but has an intact leucine zipper,48 and is only expressed in 32D cl3 cells in response to cytotoxic stress, where it dominantly inhibits C/EBP activities and reduces G-CSF responsiveness, perhaps via interference with G-CSF receptor expression.49
PU.1 (Figure 1) is a member of the Ets family of transcription factors, possessing a glutamine-rich trans-activating domain at its N-terminus, and a C-terminal DNA binding domain which binds to a consensus site 5′-AAAG(A/C/G)GGAAG-3′.5051 PU.1 is phosphorylated on multiple residues, including serines 41, 45, and 132 and/or 133, and 148.52 Both ERK1, a MAP kinase, and JNK1, a Jun kinase, can phosphorylate PU.1,53 with phosphorylation of serines 41 and 45 possibly regulating PU.1 activities in myeloid cells.54 Phosphorylation of serine 148 in PU.1 mediates its binding to Pip, an essential cofactor in B cells.55 However, the role of serine phosphorylation at this site in myeloid cells is controversial. Although phosphorylation of serine 148 was required for LPS-mediated gene activation in macrophages, mutation of this residue did not alter the ability of PU.1 to stimulate macrophage proliferation.5456 Both the glutamine-rich and an acidic domain of PU.1 show transcriptional activity in HeLa cells.57 However, these domains have only weak activity, suggesting that PU.1 needs to cooperate with additional factors to effectively contribute to gene activation. Integrity of both the C-terminal PU.1 DNA-binding domain and its N-terminal glutamine-rich trans-activating domain, but not the acidic domain, were required for PU.1 to rescue myelopoiesis in PU.1−/− ES cells.58
PU.1 is expressed in granulocytic, monocytic and B lymphoid cells.5059 PU.1 levels increase during the granulocytic differentiation of CD34+ bone marrow cells as well as the FDCP-mix and 32D cl3 cell lines.9596061 Introduction of a PU.1-ER fusion protein into MEP cells, and subsequent treatment with estradiol to activate PU.1, resulted in increased expression of two myeloid markers and decreased expression of GATA-1 and two progenitor markers.62 Similarly, activation of a PU.1-ER(T) fusion by 4-hydroxytamoxifen (4HT) in 32D cl3 cells led to increased MPO RNA expression although, in contrast to C/EBPα, terminal granulocytic maturation was not observed.35 These results are consistent with the phenotype of 32D cl3 cells stably over-expressing PU.1, which proliferate as myeloblasts with IL-3, but show accelerated differentiation and reduced proliferation with G-CSF.63 The two reported PU.1 deficient murine lines have in common a lack of monocytes and B lymphoid cells and a greatly impaired production of neutrophils.6465 Those neutrophils produced do not express some markers of terminal differentiation.66 The hematopoietic progenitors from these mice express very low levels of the receptors for M-CSF, G-CSF, and GM-CSF, and do not proliferate in response to the corresponding cytokines alone.6768 However, PU.1-deficient progenitors can produce immature myeloid cells in response to the combination of IL-3 and G-CSF,67 although introduction of the G-CSF-R or M-CSF-R into PU.1-deficient marrow cells did not enable terminal neutrophilic or monocytic differentiation in response to the corresponding factors, while introduction of exogenous PU.1 did.69 Hematopoietic colonies grown from PU.1 knockout mice in the presence of IL-3, SCF, GM-CSF and M-CSF contain cells with markers of immature monocytes.70 Thus, PU.1-deficient cells can commit to the neutrophilic and monocytic lineages, but cannot fully mature along these lineages. In contrast, yolk sac-derived phagocytes were recently shown to be distinct from embryonic and adult phagocytes and to be capable of developing in PU.1-null mice.71
Core binding factor (CBF)
CBF is a family of heterodimeric proteins containing one of three CBFα subunits bound to a common CBFβ subunit72737475767778 (Figure 1). The CBFα subunits contain a DNA-binding domain (DBD) homologous to that of the Drosophila Runt protein,79 which mediates both DNA-binding and heterodimerization with CBFβ.7680 The DNA consensus binding site for CBF is 5′-PuACCPuCA-3′.728081 Interestingly, the Runt domain can bind ATP, although the significance of this binding is not known.8283 CBFβ does not bind DNA itself, but increases the binding affinity of the CBFα subunits.7577
CBF alone has only a weak trans-activation domain.284 However, CBF potentiates trans-activation by C/EBPα, c-Myb, PU.1 and Ets-1, either via cooperative DNA-binding or via interactions with co-activators.1085868788899091 Recently, a novel mechanism of cooperativity was identified – binding of Ets-1 to CBFα2 disrupts the interaction of the CBFα2 C-terminus with the Runt domain, thereby stimulating DNA binding.92 Drosophila Runt can also repress transcription via interaction of a C-terminal motif, VWRPY, with Groucho.93 CBFα subunits similarly interact with TLE, a mammalian Groucho homologue.9495 Repression by CBFα2 may also be mediated by the mSin3 or N-CoR co-repressors.9697 CBFα2 can be phosphorylated by ERK at serines 276 and 293 and is phosphorylated in hematopoietic cells, although the role of CBF phosphorylation is uncertain.9899100 In addition, the CBFα subunits each contain a 30 amino acid segment which mediates strong affinity for the nuclear matrix.101 However, neither the nuclear matrix binding partner nor the significance of this interaction is known.
Of the three CBFα subunits, only CBFα2 (also called AML1) is expressed specifically in hematopoietic cells, being largely restricted to myeloid and lymphoid cells in adult mammals.99100102 Expression in hematopoietic stem cells has also been demonstrated using mice carrying the lacZ gene embedded in the CBFα2 locus.103 Consistent with their expression in pluripotent marrow cells, mice lacking CBFα2 or CBFβ lack all lineages of definitive hematopoiesis.104105106107108 Lack of hematopoiesis might result from the inability of primitive mesodermal cells to commit to hematopoiesis – alternatively, inhibition of cell proliferation due to loss of CBF activity, as observed in hematopoietic cell lines, may explain the phenotypes of these knockout mice.109110
The transcription factor c-Myb contains three imperfect 50–52 amino acid repeats at its N-terminus, R1, R2 and R3 (Figure 1). The R2 and R3 domains mediate DNA contact with the consensus site 5′-(C/T)AAC(G/T)G-3′, while R1 stabilizes this interaction.111112113114 These domains contain evolutionarily conserved tryptophans every 18–19 residues which help form the hydrophobic core of the helix–turn–helix DNA-binding structure.115 Two mammalian c-Myb homologues, A-Myb and B-Myb, have been identified. These three mammalian Myb family members are highly homologous in their DNA-binding domains, as well as in several other regions, with each able to bind the c-Myb consensus site.116117 c-Myb contains a centrally located trans-activating domain.118119120 The C-terminal region of c-Myb contains an evolutionarily conserved EVES motif which acts as a negative regulatory domain by interacting with the c-Myb N-terminal DNA-binding domain.121 Interestingly, a co-activator protein, p100, contains a similar EVES motif and can also interact with the c-Myb DNA-binding domain.121 Whether p100 activates or represses c-Myb activity in vivo by this interaction remains to be determined. Phosphorylation of serines 11 and 12 of c-Myb, by casein kinase II in vitro or by unidentified kinases in vivo, inhibits c-Myb DNA-binding.122 Phosphorylation of a serine just downstream of the EVES motif reduces c-Myb's trans-activation potency.123 Perhaps this phosphorylation stabilizes the interaction of the EVES domain with the c-Myb N-terminus. In addition to the EVES motif, the negative regulatory domain of c-Myb contains a leucine zipper. Mutations in this motif increase c-Myb DNA affinity, and phosphorylation of serines 11 and 12 prevents leucine zipper-mediated inhibition of DNA binding.124125 A protein, termed p160, has been identified which interacts with c-Myb via its leucine zipper.126 Perhaps this protein also interacts with the c-Myb N-terminus and inhibits DNA-binding if serines 11 and 12 are not phosphorylated.
c-Myb is predominantly expressed in immature lymphoid, erythroid, and myeloid cells and has also been detected in cells of the intestinal crypt and hair follicle in adult mice.127128129130 Consistent with its expression pattern, c-Myb knockout mice lack all hematopoietic lineages except the megakaryocytic branch.131 In addition, ectopic expression of c-Myb in 32D cl3 cells allowed G-CSF induction of early differentiation markers, such as MPO. However, terminal differentiation, with associated growth arrest and lactoferrin induction, was inhibited.132 These results are in agreement with the idea that c-Myb stimulates cell proliferation and that cell cycle arrest is required for the terminal phases of granulopoiesis. Indeed c-Myb has been implicated in the regulation of the G1/S cell cycle transition, and Drosophila Myb has been shown to be required for the G2/M transition.133134135 A-Myb is expressed in the developing CNS, in the adult testis, and in a subset of germinal center B lymphoid cells,117136137 while B-Myb is expressed ubiquitously in proliferating cells and may help regulate cell cycle progression.138139 The severe hematopoietic phenotype present in c-Myb-null mice indicates that neither A-Myb nor B-Myb can compensate for the loss of c-Myb in early hematopoietic cells. Whether Myb family members have overlapping roles in subsets of hematopoietic cells remains to be determined.
Homeobox (Hox) proteins
Hox proteins possess a specific helix–turn–helix DNA-binding motif, termed the homeo domain.140 Multiple members of this protein family are expressed in hematopoietic cells, with a number of them specifically implicated in granulopoiesis. HoxB4 is expressed in the most early hematopoietic progenitors. Overexpression of this protein leads to a selective expansion of these primitive progenitors without affecting their differentiation.141 Both HoxA9 and HoxA10 are highly expressed in human CD34+ cells, but are down-regulated during myeloid development.19142 Hematopoietic analysis of mice with a targeted disruption of the hoxA9 gene revealed reduced granulocytes (as well as lymphocytes), and a decreased responsiveness to G-CSF, suggesting that HoxA9 has an important function in committed progenitor cells.142 In contrast, hoxA10 knockouts showed a two-fold increase in circulating neutrophils (and monocytes).143 In addition, overexpression of HoxA10 in mouse bone marrow perturbed myeloid differentiation and led to acute myeloid leukemia.141 HoxA5 is also expressed during early myelopoiesis – inhibition of HoxA5 expression with antisense oligonucleotides inhibited granulocytic and monocytic hematopoiesis.144 Other homeobox genes, such as Hlx and HoxB7, are expressed in mature granulocytic cells.145146 Retroviral expression of Hlx enhanced myeloid maturation,145 while overexpression of HoxB7 inhibited granulocytic differentiation of HL-60 cells.146 Thus, specific Hox family members appear to act as important regulators at distinct stages of lineage commitment and granulopoiesis.
Wilms’ tumor suppressor protein (WT-1)
The WT1 protein contains four zinc fingers at its carboxyl terminus and a transcriptional regulatory domain.147 Within the coding region of this transcriptional regulatory domain, exon 5 is alternatively spliced,148 while RNA editing occurs within exon 6 which can modify its activity.149 A second alternative splicing event inserts or removes a three amino acid sequence (KTS) from the 5′ end of exon 10.148 The KTS insert disrupts the spacing between zinc fingers 3 and 4 which profoundly affects DNA binding. Isoforms lacking the KTS insert bind to a GC-rich motif that is very similar to the known Egr-1 binding site, suggesting that WT1 possibly modulates the ability of Egr-1 to induce monocytic differentiation.150 Other high-affinity binding sites for WT1 have also been identified.151 Although isoforms containing the KTS insert bind DNA, a consensus binding site has not been established.152 WT1 can mediate both transcriptional activation and repression.153154155 WT1 isoforms with the KTS insert interact with components of the RNA splicing machinery in vitro,156157 and in living cells they are localized to nuclear coiled bodies, which are hypothesized to play a role in the assembly and maturation of splicing complexes.156
WT1 is expressed in CD34+ bone marrow cells20158 – treatment of these cells with stem cell factor and G-CSF down-regulates its expression.159 WT1 expression is similarly reduced in HL-60 and K562 cells induced to differentiate.160161 In M1 cells, WT1 expression is induced by treatment with leukemia inhibitory factor. Exogenous expression of isoforms containing the KTS insert induces spontaneous monocytic differentiation of these cells, while expression of isoforms lacking the KTS insert induces G1 arrest and apoptosis.162 In contrast, enforced expression of the isoform containing both exon 5 and the KTS insert in 32D cl3 cells blocks G-CSF-mediated granulocytic differentiation and promotes survival in the absence of IL-3,163 while expression of the variant lacking exon 5 and the KTS insert accelerates G-CSF-induced differentiation (DML, ADF and S Sukumar, unpublished). Perhaps the opposing effects on monocytic and neutrophilic differentiation of WT1 isoforms with or without the KTS insert reflects the action of WT1 during normal hematopoiesis.
Several target genes of WT1 have been identified that may be important in the control of renal or gonadal differentiation, such as the EGF receptor, the insulin-like growth factor receptor, and Dax-1.164165166 We recently found the WT1 isoform lacking both exon 5 and the KTS insert binds to sites within the cyclin E promoter, down-regulates its transcriptional activity, and reduces cyclin E protein expression in 32D cl3 cells (DML ADF and S Sukumar, unpublished), which possibly contributes to differentiation induction. WT1 also upregulates the anti-apoptotic gene bcl-2,167 which is of considerable interest in view of the finding that overexpression of WT1 is associated with a poor outcome in patients with acute leukemia.21 Finally, repression of the human RARα gene via a GC-rich sequence may be relevant to regulation of hematopoiesis by WT1.168
Retinoblastoma protein (Rb)
When hypophosphorylated, Rb cooperates with E2F family members to inhibit expression of genes required for S-phase entry.169 Rb knockout mice show defective hematopoiesis with an increased number of immature erythrocytes entry.170 Rb expression is down-modulated in granulocytic differentiation, while it is up-regulated during erythroid and monocytic differentiation.171172 Suppression of Rb expression in human marrow myeloid progenitors using anti-sense oligonucleotides inhibited monocytic in favor of granulocytic differentiation without effecting their colony-forming ability,172 suggesting that Rb functions as a switch between the monocytic and granulocytic lineages. The ability of Rb to interact with C/EBPβ and PU.1 may be of relevance to this observation.173174 When complexed to these factors, hypophosphorylated Rb may potentiate their ability to contribute to both monocytic and neutrophilic differentiation, with one of these complexes being in fact essential for monopoiesis.
Retinoic acid receptor (RAR)
Retinoids have long been known to participate in myeloid differentiation, with addition of retinoids observed to stimulate granulopoiesis,175176 while lack of retinoids (for example, vitamin A deficiency) has been associated with defective hematopoiesis.177 In addition, retinoic acid could induce granulocytic differentiation of the HL60 myeloid cell line and of primary cells from acute promyelocytic leukemia (APL) patients.178179 Retinoid activity is mediated via two different families of nuclear receptors: the retinoic acid receptors (RARα, RARβ, RARγ) which bind both all-trans retinoic acid (ATRA) and 9-cis retinoic acid,180 and the retinoid X receptors (RXRα, RXRβ, RXRγ) which only bind 9-cis retinoic acid.181 Of the three RARs, RARα is preferentially expressed in myeloid cells.182
RARα binds to specific DNA sequences called retinoic acid response elements (RAREs), which consist of a direct repeat of the sequence (A/G)G(G/T)TCA separated by 2–5 nucleotides.183184 High-affinity binding of RARs is achieved via heterodimerization with a member of the RXR family, and is regulated by the presence or absence of ligand. In the absence of ATRA, RARα is associated with the co-repressor proteins N-CoR and SMRT. This leads to recruitment of transcriptional repressor proteins, such as mSin3, as well as histone deacetylases (HD).185186 Subsequent deacetylation of histone proteins results in nucleosome assembly and transcriptional repression, as the chromatin becomes inaccessible to both transcriptional activators and the basal transcription machinery.187188 In the presence of ligand, the corepressor and HD proteins are released and replaced by coactivator proteins, such as TIF1α, SRC-1 and ACTR.189190191 This leads to acetylation of the histone proteins – either directly through co-activators like SRC-1 and ACTR, or indirectly through recruitment of CBP and p/CAF – resulting in transactivation of gene expression.191192
While treatment of cells with ATRA ultimately induces many genes, including cytokine receptors, enzymes and adhesion molecules, relatively few appear to be direct target genes. Indeed, many of the direct targets are themselves transcription factors, including C/EBPε, Hox proteins and Stat1.193194195 However, in myeloid cells RARα induces the expression of the p21waf1/cip1 cyclin-dependent kinase inhibitor through interaction with an imperfect RARE in the p21 promoter,196 as well as E3, a hematopoietic specific factor of unknown function.197
Expression of a dominant-negative RARα in a multipotent cell line changed the developmental fate of these cells from the granulocytic/monocytic to the mast cell lineage.198 GM-CSF-induced differentiation of these cells was blocked at the promyelocyte stage.199 In addition, while no major perturbations in hematopoiesis were observed in RARα knockout mice,200 RARα−/− ES cells fail to mature into granulocytes.201 More recent studies with purified cell subsets have shown that ATRA delays the maturation of primitive hematopoietic precursors, while enhancing the terminal differentiation of committed granulocytic/monocytic precursor cells,202 which might partially explain this anomaly.
Stats are latent cytoplasmic transcription factors that are recruited to activated cell-surface receptors, where they become tyrosine phosphorylated. Subsequently, these Stats homo- or hetero-dimerise, translocate to the nucleus, and effect gene expression by binding to specific promoter elements. Seven Stat proteins have been identified in mammalian cells, with additional isoforms of several Stats also identified.203 Individual Stats bind to similar DNA response elements, variants of the sequence TTN4–6AA.204205206 However, there is a considerable degree of specificity, such that different genes are targeted by different Stat complexes.207 Stat proteins interact with a number of other nuclear factors and co-activators, including CBP, Nmi, the glucocorticoid receptor, c-Jun and MCM5,203208209210 which increases the scope of transcriptional responses in which Stats can participate. While tyrosine phosphorylation is critical for Stat activation, serine phosphorylation probably also affects their activity.211212213214 Indeed, Ser727 of Stat1α has been shown to be directly involved in the recruitment of MCM5 as part of interferon γ-induced transcriptional activation.209 Several important genes are activated by Stat proteins, including p21WAF1/Cip1, c-fos, cis and cyclin D1.215216217218
Cytokines important for granulopoiesis are known to activate Stats. GM-CSF and IL-3 activate Stat5,219 while G-CSF activates Stat3, Stat5 and some Stat1.220221222223 Targeted disruption of the Stat3 gene gave early embryonic lethality,224 while mice with both forms of Stat5 knocked out exhibited multiple defects, with responses to IL-3, GM-CSF and G-CSF affected.225 Expression of dominant-negative Stats, or specific receptor mutants, has provided further evidence of the essential role of Stats in mediating responses from cytokine receptors during granulopoiesis. These studies have shown, for example, that Stat3 activation plays a key role in the differentiation responses to G-CSF,226227228 while Stat5 appears important for proliferative responses to IL-3, GM-CSF, and G-CSF.219229230
The c-myc gene encodes a basic helix–loop–helix/leucine zipper-containing transcriptional activator protein c-Myc. This dimerises with another small basic helix–loop–helix protein Max in order to bind DNA and transactivate gene expression from E-box (CACGTG) sites.231232 Myc expression is rapidly induced following cytokine stimulation of responsive cells.233234 Many known c-Myc target genes are involved in cell cycle progression or protein translation.235 Consistent with this, the c-myc gene is known to have oncogenic potential and high expression of c-myc is associated with proliferation and a block in differentiation.236237 Both the oncogenic activity and the cell proliferation promoting activity is mediated by the Myc-Max heterodimeric complexes.238 How c-Myc controls the cell cycle is not completely understood. Apart from the transcriptional control of target genes directly involved in DNA synthesis or protein translation (such as ornithine decarboxylase), two more direct cell cycle modifying effects of c-Myc have been suggested: (1) via upregulation of cdc25A and thereby activation of cyclin-dependent kinases, and (2) via sequestration of the general G1 cyclin-dependent kinase inhibitor p27Kip1 from G1 cyclin/CDK complexes.239240 The latter mechanism would be mediated by yet unidentified c-Myc target gene(s), the protein product of which can probably directly bind to p27. Of note, evidence suggesting that cdk4 can sequester p27 was recently presented.241 Studies in a c-Myc knockout cell line further establish a role for c-Myc in the control of cell proliferation. In these cells, the kinase activities of all cyclin/cdk complexes were reduced, while p27 levels were increased, leading to a concomitant delay in cell cycle entry.242
Mad1 belongs to a family of small basic helix–loop–helix/leucine zipper-containing transcription repressor proteins which includes Mad1, Mxi,243 Mad3 and Mad4.244 All four proteins can dimerise to the same Max protein as c-Myc and compete for binding to the same E-box sequences as c-Myc/Max heterodimers. Mad1 had been shown to counteract c-Myc activity and inhibit cell proliferation through its transcriptional repressor activity.245246 This lies in its N-terminal mSin3 binding domain (SID)245247 which recruits HD proteins via binding of mSin3.246248 Expression of mad1 and mxi in hematopoietic cell lines has been reported,249250251 whereas mad3 and mad4 seem to be more restricted to epithelial and neural tissues.244 mad1 was shown to be induced during differentiation of several hematopoietic lineages in vitro249250251 and in vivo.252 Indeed, the differentiation of many cell types is associated with opposite regulation of c-myc (down-regulation) and mad1 (up-regulation). That Mad1 may play an important role in the differentiation of myeloid lineages is further supported by data from mad1 knockout mice that exhibited a delayed cell cycle exit of the granulocytic lineage.253
Apart from those described above, a number of other transcription factors appear to play a role in granulocytic cells, but these will not be discussed in detail here. Briefly, these include: CDP: The ubiquitous CCAAT displacement protein (CDP) is a negative regulator of a subset of genes which are expressed specifically in more mature granulocytes, including those encoding gp91-phox and lactoferrin.254255256 Sp1: Another widely expressed factor, Sp1 is present at particularly high levels in maturing granulocytes and has been implicated in the regulation of several myeloid genes.14 MZF-1: A zinc-finger protein expressed specifically in early myeloid cells, MZF-1 is required for the formation of myeloid colony-forming units.257 Its enforced expression in a myelomonocytic cell line delayed differentiation and augmented survival.258 Id proteins: In hematopoietic cells, the inhibitory helix-loop-helix proteins Id1 and Id2 are expressed reciprocally during growth and differentiation, with Id2 expression observed to increase as cells mature.259260 Id protein levels decrease early when 32D cl3 cells are exposed to G-CSF, with enforced expression of Id in these cells leading to apoptosis, rather that differentiation, in response to G-CSF.261 PLZF: The zinc-finger transcription factor PLZF is expressed in immature but not more mature hematopoietic cells. Expression of exogenous PLZF in 32D cl3 myeloid cells slowed their proliferation in IL-3 and inhibited their differentiation in G-CSF.262 The inhibition of proliferation may have resulted from direct repression of cyclin A gene transcription.263
Transcriptional control of lineage commitment
Figure 2 presents a model of the transcriptional regulation of lineage commitment within hematopoiesis. Implicit in this model is the concept that transcription factors rather than cytokine receptor signals are the key determinants of lineage commitment. Consistent with this idea, mice lacking G-CSF, the G-CSF receptor, or G-CSF and GM-CSF have at most a two-fold reduction in immature granulocytic precursors and bone marrow neutrophils,264265 whereas those lacking C/EBPα or PU.1, for example, show a much greater impairment in granulocyte production.366465 Furthermore, introduction of the G-CSF-R into PU.1-deficient bone marrow cells failed to restore terminal granulocytic differentiation of these cells.69
The helix–loop–helix transcription factor SCL has been portrayed as the master regulator of hematopoiesis, being absolutely required for generation of a pluripotent stem cell.266 Subsequent development of pluripotent hematopoietic stem cells requires CBF, c-Myb and GATA-2. The recent observation that GATA-2 can bias murine marrow cells and FDCP-mix cells toward granulocytic differentiation raises the possibility that GATA-2 regulates C/EBPα levels.267 Induction of bipotent lymphoid progenitors requires Ikaros and PU.1. GATA-3 is required for T lymphoid development, while both PU.1 and Pax-5 contribute to B lymphoid development. GATA-1 in combination with C/EBPα induces the eosinophil lineage while, in the absence of C/EBPα, GATA-1 induces the erythroid/megakaryocyte lineages.38 GATA-1 in combination with Friend of GATA (FOG) and NF-E2 subsequently facilitates megakaryocytic differentiation, while GATA-1 in combination with erythroid Krüppel-like factor (EKLF) facilitates erythroid development. Finally, PU.1 and C/EBPα are important for myeloid development. In this model, C/EBPα induces the granulocytic/monocytic lineages, with induction of PU.1 to high levels. This model of hematopoiesis takes account of the finding that C/EBPα activates PU.1 gene expression35 thereby placing C/EBPα upstream of PU.1 in myeloid development – with both cooperating in granulocytic differentiation, but with PU.1 being more important in monocytic differentiation. A question remains open in this model – namely, what determines whether a bipotential G/M progenitor commits to the granulocytic or monocytic lineage? Perhaps this choice is governed by C/EBP levels.
Transcription factor cooperativity in granulopoiesis
While no single transcription factor is expressed only in early myeloid cells (monocyte and granulocyte precursors), the combination of C/EBPα with PU.1, CBF and c-Myb is unique to these lineages. Indeed, it appears that it is the cooperation between these factors – together and with other factors – which is the major driving force in mediating granulopoiesis. C/EBPβ and c-Myb cooperate to activate the endogenous mim-1 gene in avian cells.1 C/EBPα, c-Myb and PU.1 cooperate to activate the murine neutrophil elastase (NE) promoter, and probably cooperate to activate the highly homologous proteinase-3 and azurocidin genes as well,39 while other C/EBPs, such as C/EBPβ and C/EBPδ, and other Ets factors, such as GABP and Ets-2, can also cooperate with c-Myb to activate the NE promoter.268 In addition, the myeloperoxidase (MPO) gene contains two enhancers, a proximal enhancer regulated synergistically by CBF and c-Myb and a distal enhancer activated by PU.1 and C/EBPα.1189 Importantly for understanding the hierarchy of granulopoietic inducers, both the G-CSF receptor and GM-CSF receptor promoters are regulated by PU.1 and C/EBPα,712 and the M-CSF receptor promoter is activated synergistically by C/EBPα, CBF and PU.1.41090
Transcription inhibitors also contribute to control myeloid gene expression and lineage determination. For example, GATA-1 interacts with PU.1 and interferes with its activation of myeloid genes in erythroid cells.269270 MafB inhibits Ets-1-mediated activation of erythroid genes in myeloid cells, and a related bZIP factor, c-Maf, inhibits Ets-1 and c-Myb activation of the CD13 gene in early myeloid cells which have not yet reached the stage of differentiation at which CD13 expression becomes evident.271272 Inhibition of c-Myb activities is not the only functionally important activity of c-Maf, as c-Maf overexpression directed both the HL-60 and U937 myeloblastic leukemia cell lines to differentiate to monocytes, whereas a dominant-negative c-Myb did not.273
The transcription factors described above can also influence each other's expression. For example, C/EBPα-ER rapidly induces PU.1 mRNA levels, even in the presence of cycloheximide in both 32D cl3 myeloid cells and Ba/F3 B lymphoid cells,35 suggesting that C/EBPα acts directly on the PU.1 promoter. In addition, both C/EBPα and PU.1 can directly auto-activate their own promoters.274275276 Furthermore, Myb regulates the expression of c-Myc,277 and co-operates with GATA-1 to enhance WT-1 expression.278
Cytokine receptor signals and granulopoiesis
Cytokine receptors expressed in cells of the granulocytic lineage
The production of granulocytic cells is also regulated by a network of hematopoietic growth factors and cytokines. One of these, granulocyte colony-stimulating factor (G-CSF), is a major regulator of neutrophilic granulocyte production, and augments the proliferation, survival, maturation and functional activation of cells of this lineage.264265 The actions of G-CSF are mediated through its interaction with a specific cell-surface receptor, the G-CSF-R, which forms homo–oligomeric complexes upon ligand binding.279 Typical of other members of the hematopoietin receptor superfamily, the G-CSF-R has no intrinsic tyrosine kinase activity but activates cytoplasmic tyrosine kinases.279280 In addition, the cytokines IL-3 and GM-CSF also play important roles in granulocytic cells through signals from their specific receptors, which have been reviewed recently.281282
The G-CSF-R is an 813 amino acid membrane protein, including a 188 amino acid cytoplasmic segment which is diagrammed in Figure 3. The cytoplasmic region of the G-CSF-R can be subdivided into a membrane-proximal domain, which contains two conserved sequences known as box 1 and box 2, and a membrane-distal domain, which contains a less-conserved box 3 sequence.279 In myeloid cells, the membrane-proximal domain is essential for mitogenic signaling, while the membrane-distal domain is essential for the transduction of differentiation signals.223283284285 In addition, there are four tyrosine (Y) residues in the cytoplasmic region of the G-CSF-R, at positions 704, 729, 744 and 764 of the human receptor, three of which lie in the membrane-distal domain.286 Ligation of the G-CSF-R results in the rapid phosphorylation of these four tyrosines,287288 which create binding sites for signaling molecules. Some signaling pathways emanating from the different tyrosines of the G-CSF-R have been identified. For example, we and others have shown that Y704 and Y744, and to a lesser extent Y729, of the G-CSF-R are involved in the recruitment and activation of Stat3,220222228289290 while SHP2/Grb2 complexes dock at Y704 and Y764.228 In addition, Y764 is necessary for the formation of Shc/Grb2/p140 and Grb2/p90 complexes, as well as the activation of p21ras.228234291 Jak2 and/or other Jak kinases bind to the membrane proximal segment between box 1 and box 2 – this binding is dependent upon the integrity of tryptophan 650 (W650).292 Jak2 activates other proteins, including Stat1, Stat5 and c-Rel.223293294 The G-CSF-R also activates the src-kinases Lyn and Hck,295296297 the Syk kinase,295296 as well as phospho-inositol 3′-kinase (PI3K), which in turn activates PKB/Akt.298 Interestingly, mice lacking SHIP, a phosphatase which hydrolyzes the product of PI3K have elevated PKB/Akt kinase activity and myeloid hyperplasia.299
To better define the role of signaling pathways from the G-CSF-R, we recently examined the ability of a series of tyrosine mutants to transduce biological signals in response to G-CSF in maturation-competent myeloid 32D cells.228 These cells can closely recapitulate many of the normal responses to G-CSF, including proliferation, survival, and, importantly, terminal differentiation into mature neutrophils,285 and therefore provide an appropriate cellular context to assess the physiological relevance of these signals. This analysis revealed that multiple tyrosines contribute to proliferation, differentiation and survival signaling from the human G-CSF-R. Analysis of signaling pathways downstream from these tyrosines suggest a positive role for Stat3 activation in both differentiation and survival,228 in agreement with data obtained using dominant-negative Stat3 in myeloid cells.227300301 In contrast, SHP-2, Grb2 and Shc appear to be important for proliferation,228 suggesting an important role for p21ras, which lies downstream.234 Consistent with this conclusion, a dominant-negative Ras, RasN17, inhibited the proliferation, but not the differentiation, of 32D cl3 cells.302 Finally, we also showed that an as yet ill-defined tyrosine-independent ‘differentiation domain’ in the membrane-distal region of the G-CSF-R appears necessary, although not sufficient, for mediating neutrophilic differentiation in 32D cells.228
Role of G-CSF-R signals in granulopoiesis
Mice lacking G-CSF, the G-CSF-R, or G-CSF in combination with GM-CSF have at most a two-fold reduction in immature granulocytic precursors and marrow neutrophils.264265 Peripheral neutrophils are reduced approximately five-fold, perhaps due in part to a defect in their release from the marrow.303 Redundant signals from other cytokine receptors may compensate for the lack of G-CSF-R signals in these mice. Additional evidence for a direct role of G-CSF-R signals in granulopoiesis comes from patients with severe congenital neutropenia (SCN). Their marrow shows an arrest at the promyelocyte or myelocyte stages of neutrophil development. Truncations of the G-CSF-R cytoplasmic domain have been detected in approximately 25% of SCN patients.283304 Mice carrying such a mutation are also neutropenic, although not as severely as human patients with SCN.305 In addition, we have recently identified a mutation in the extracellular domain of the G-CSF-R in another SCN patient.230
Loss of signals from the M-CSF receptor in mice lacking M-CSF can be compensated for by transgenic expression of an anti-apoptotic protein, bcl-2, in monocytic cells.306 Similarly, transgenic expression of bcl-2 in T cells rescues T lymphocyte development in mice lacking the IL-7 receptor.307 However, bcl-2 does not rescue erythroid development in mice lacking the erythropoietin receptor.308 Similarly, bcl-2 did not enable 32D cl3 cells to express MPO or Cathepsin G in the absence of G-CSF, although nuclear morphologic changes did occur.309 Thus, some receptors, including the G-CSF-R, probably provide more than merely survival signals to enable lineage development.
Transcriptional collaboration with cytokine receptor signals
It is presumed that the bulk of signals emanating from cytokine receptors ultimately converge in the nucleus to affect myeloid gene transcription, largely via stimulation, suppression and collaboration with a range of transcription factors. Cytokine receptors, including the G-CSF-R, activate c-Jun via the JNK pathway.310 c-Jun and PU.1 physically interact via their DNA-binding domains and have been shown to cooperate to activate transcription in U937 cells via a PU.1-binding site.311312 c-Jun and PU.1 also cooperate to activate the macrosialin gene in monocytic cells via neighboring AP-1 and PU.1 DNA-binding sites.313 In addition, the receptors for G-CSF, GM-CSF and IL-3 all activate Ras and components of the MAPK pathway.234282 MAPKs are known to phosphorylate a number of transcription factors, including PU.153 and C/EBPβ.314 In addition, Ras and C/EBPα cooperatively activate the G-CSF-R promoter in 293T cells, an adherent epithelial cell line.315 This activation is dependent upon phosphorylation of serine 248 by protein kinase C (PKC). Whether Ras activation leads to PKC-dependent C/EBPα phosphorylation in hematopoietic cells is unknown. In addition, Ras activation is required for c-Myb to induce GBX2, a homeodomain protein, in avian myeloid cells.316 The c-Myb interacting protein p100 can be activated by the Pim-1 kinase, which itself is activated by Ras, potentially providing the link to explain the involvement of Ras in modulating c-Myb activity.317
Typical of other cytokine receptors, the G-CSF-R induces c-Myc,234 which itself induces important cell cycle genes, as well as indirectly sequestering p27 to increase cyclinE/CDK2 activity.239240 In addition, Stat5 is activated by the G-CSF-R, GM-CSF-R and IL-3-R219223 which can induce cyclin gene expression218318 to further promote proliferation in the early phases of granulocytic differentiation. In the later phases of G-CSF-mediated granulocytic differentiation, Stat5 activation decreases, while Stat3 remains high.223 This leads to increased expression of p27 thereby promoting growth arrest.301 Consistent with a role for Stat3 in delayed growth arrest, dominant-negative Stat3 prevents terminal granulopoiesis and enables the continuous proliferation of LGM-1(G-CSF-R) and 32D cl3 cells in the presence of G-CSF.227301 Mad1 is also induced during G-CSF-mediated granulopoiesis which can inhibit Myc activities (AS-B, D van Leeuwen, Y van Aesch, IPT, unpublished data). Such a synergistic effect of Mad1 and p27 in achieving growth arrest in granulocytic cells has also been reported using murine knockout models.253 Potentially related to these observations, C/EBPα, like G-CSF, induces a delayed G1/S cell cycle arrest in 32D cl3 cells.35 Induction of cell cycle arrest was independent of differentiation, as it occurred also in the presence of p210Bcr-Abl, which prevented all aspects of C/EBPα-induced granulopoiesis except PU.1 induction. C/EBPα also induces cell cycle arrest in adipocytes and hepatocytes, perhaps via direct interaction with p107, a homologue of Rb.319320321 Induction of p27Kip1 was also evident in 32D cl3 cells expressing C/EBPα-ER.35 Therefore, we would speculate that Stat3 activation of the p27 promoter301 occurs in cooperation with a C/EBP or another transcription factor active in maturing granulocytes. Either induction of p27 or direct inactivation of an Rb or p107 complex could account for the recent finding that C/EBPα inhibits c-Myc gene transcription in a myeloid cell line via an E2F-binding site in its promoter.322
C/EBPα:G-CSF-R interactions – a model of granulopoiesis
We have recently obtained evidence that G-CSF-R signals, although redundant with signals of other cytokine receptors, can cooperate with exogenous C/EBPα to activate the endogenous MPO and NE genes in a lymphoid cell line.323 In addition, the integrity of both the C/EBPα- and PU.1-binding sites is known to be required for induction of the neutrophil elastase promoter by G-CSF receptor signals,39 while NF-Y may play a role in G-CSF-dependent MPO transcription, via its binding site in the MPO promoter.324 Perhaps G-CSF-R signals contribute to C/EBPα gene induction in immature myeloid cells, as C/EBPα protein and RNA levels and transcription initiation rate increase rapidly when 32D cl3 cells are transferred from IL-3 to G-CSF31 (L Scott and ADF, unpublished). C/EBPα – in co-operation with PU.1 – in turn activates the G-CSF-R promoter, and perhaps the IL-6 gene, generating a positive feedback loop which fixes the commitment of myeloid progenitors.12325
Figure 4 presents a model of granulopoiesis beginning with G-CSF receptor signals. While these signals may be redundant with those emanating from other receptors with respect to the regulation of early granulopoiesis, they may be required specifically for late granulopoiesis,265 where they cooperate to elicit cell cycle arrest and induce specific genes, just as signals from the TPO receptor are required for terminal thrombopoiesis.326327 Alternatively, the major function of G-CSF-R signals may be to provide an enhancement of granulocyte production and survival required, for example, in response to infection when G-CSF levels rise steeply.264 In this scenario, induction of Myc and activation of Stat5 could clearly be important for inducing proliferation.
Studies of gene regulation in immature myeloid cells have pinpointed a group of transcription factors which coordinate early granulopoiesis. Similar investigations of genes transcribed specifically during late granulopoiesis are beginning to uncover factors which control this phase of differentiation. Additional transcription factors have also been identified based on their lineage-restricted expression pattern or their role in leukemogenesis as being important for regulating granulopoiesis. A key goal for future investigation is to organize these various factors and signals emanating from cytokine receptors into pathways which direct each step of maturation, from the pluripotent hematopoietic stem cell, to a committed myeloid progenitor, to myeloid precursors, and ultimately to mature granulocytes. This goal will be facilitated by identifying the primary genetic targets of each factor, the phosphorylation sites which regulate their activities, and the co-activators which facilitate their function. Finally, thorough characterization of mice with lineage- and stage-specific perturbations in the expression of one or more regulatory factor will also assist in the development of a molecular model of granulopoiesis.