Myeloid-specific gene expression

Transcription

1 Myeloid-specific gene expression Sandra Clarke and Siamon Gordon Sir William Dunn School of Pathology, University of Oxford, United Kingdom Abstract: The mononuclear phagocytes are recruited from bone marrow precursors to most tissues of the body, particularly during inflammation or immune stimulation. This combination of accessibility as stem cells and heterogenity of tissue locations makes the myeloid cell potentially important as a carrier of therapeutic agents. Understanding the regulation of transcription in myeloid cells is necessary for any future design of tissue-specific gene targeting vectors, particularly if there are inherent size limitations. Identified members of the C/EBP, Runt/PEBP2/CBF, and Ets families of transcription factors are critical for myeloid-specific gene expression and may have myeloid-restricted expression or myeloid-specific regulation in the hematopoietic system. AP-1, Sp1, and Myb appear to be important for myeloid-restricted expression in some cases. In addition, factors involved in the up-regulation of the level of gene expression when macrophages are activated by agents such as interferon- and bacterial products have been identified. Some of the sequences to which these transcription factors bind in myeloid-restricted genes have been tested in cell lines and in transgenic mice and it is now possible to make an attempt to describe the characteristics of a myeloid-specific promoter. J. Leukoc. Biol. 63: ; Key Words: transcription factors gene expression macrophages INTRODUCTION The monocyte expresses a wide and variable range of myeloidrestricted genes [1 3] as it migrates from the bone marrow to join a particular subset of resident tissue macrophages, and if the cell encounters and responds to environmental stimuli in its role as a major player in primary immunological defense, the number of induced, cell-type-restricted, expressed genes will be significantly larger. At each stage of commitment from pluripotent stem cell to monocyte to tissue macrophage, or to activated macrophage, a different combination of restricted and ubiquitous regulatory factors is likely to be important in inducing the expression of a changing pattern of genes. Myeloid-restricted gene expression in granulocytes tends to be confined to the early stages of maturation in the bone marrow. At present there are more than 30 cloned genes expressed by macrophages and granulocytes (Table 1) with some claim to myeloid-restricted expression and the cis and trans elements for many of these have now been described, albeit only for positive regulation of the proximal promoter regions in most cases. In the following sections, the transcription factor families that appear to control myeloid-specific gene expression will be summarized. Following this, the isolated and characterized regulatory regions of some of the genes listed in Table 1 will be discussed, and an attempt will be made to describe the requirements for myeloid-specific gene expression. Finally, the summary of transgenic mice (and one rabbit) will show that the identified regulatory regions of some of these genes direct myeloid-specific expression in vivo. TRANSCRIPTION FACTORS INVOLVED IN MYELOID GENE EXPRESSION The transcription factor families that have been shown to be important for the transcription of myeloid-specific genes are the C/EBP family, particularly members C/EBP and C/EBP [nuclear factor-interleukin-6 (NF-IL-6)]; the Runt/PEBP2/CBF family, particularly AML1/PEBP2 B/CBF 2; the Ets family, particularly PU.1 and Ets-1 or Ets-2; the ubiquitous AP-1 group; and Sp1 (Tables 1 3). Although the Sp1 and AP-1 families appear to be critical for myeloid-specific expression of some genes (Table 1) both families are ubiquitously expressed. AP-1 mediates gene induction by numerous stimuli to which myeloid cells are responsive, such as TPA [4], tumor necrosis factor (TNF- ) [5], and granulocyte-macrophage colonystimulating factor (GM-CSF) [6], and in general is induced by a range of growth factors, cytokines, and other stimuli that activate protein kinase C [7] The single, universally expressed protein, Sp1, associated with promoters without TATA boxes [8], may acquire tissue-specific regulation through modifications such as phosphorylation and glycosylation [9, 10] or modulation of expression levels [11]. For example, the 105- kda, phosphorylated form of Sp1 is recognized by the Sp1 site of human CD14, and a human monocyte cell line has 65% less Sp1 than HeLa cells, as well as a higher ratio of the phosphorylated form [12]. The CCAAT/enhancer binding protein (C/EBP) family The capacity of C/EBP family members to bind avidly to two conserved cis sequences found in animal viruses, CCAAT Abbreviations: TNF-, tumor necrosis factor ; GM-CSF, granulocytemacrophage colony-stimulating factor; M-CSF, macrophage colony-stimulating factor; G-CSF, granulocyte colony-stimulating factor; IFN-, interferon- ; inos, inducible nitric oxide synthase; NO, nitric oxide; MIP-1, macrophage inflammatory protein-1 ; MSR, macrophage scavenger receptor; LPS, lipopolysaccharide. Correspondence: Dr. Sandra Clarke, Sir William Dunn School of Pathology, University of Oxford, Oxford OX1 3RE, United Kingdom. Received September 16, 1997; accepted November 20, Journal of Leukocyte Biology Volume 63, February

2 TABLE 1. Cloned Myeloid-Restricted Genes Gene Binding sites and response regions References Genes primarily expressed in macrophages M-CSF receptor (c-fms) PU.1,* PEBP2/CBF,* C/EBP * (h) [17, 56, 57, 91, 92] PU.1, PEBP2/CBF (m) Macrophage scavenger receptor (MSR) TPA response region of AP-1/Ets-2,* PU.1* (h) [94, 97] Lipopolysaccharide receptor (CD14) PU.1,* Myb,* Sp1,* AP-1, AP-2 (h) [12, 103] TPA, IFN-, vitamin D (m) Macrophage mannose receptor TATA, CCAAT, AP-1, AP-2, Sp1, NF- B (h) [176] Sialoadhesin [177] F4/80 [178] Macrosialin (CD68) [179, 180] Natural resistance-associated macrophage protein IFN- response regions (h, m) [170, 171, 181] (Nramp) NF-IL-6, AP-1, Sp1, PU.1, (m); NF- B (h) Visna virus AP-1* [182, 183] Monocyte specific esterase (MSE) [184, 185] Macrophage metalloelastase TATA, Ets, AP-1 (h) [186] Macrophage gene-1 (Mpg-1) [187] IFN- -inducing factor (IGIF) [188] Genes primarily expressed in both macrophages and granulocytes CD11b (CR3) Spl,* PU.1,* MS-2* (h) [ , 189, 190] Fc R1 (CD64) (GIRE-BP),* PU.1 (h) [ ] [172, 191] Fc RIII (CD16) PU.1,* MyE (myeloid E box)* (m) [117, 120] NF-IL-6, GATA, -IRE, AP-1, AP-2, Ets, Sp1 (h) GM-CSF receptor PU.1,* C/EBP * (h) [29, 192] c-fps/fes Sp1,* PU.1* (h) [ ] gp91-phox CP1/CDP,* IRF1,* IRF2,* TF1 phox* (h) [130, 131, 135, 175, 196, 197] Formyl peptide receptor (FPR) Ets, AP-1 (h) [198, 199] Lysozyme PU.1,* NF-M* (c); GABP* (m) [25, 151, 169, ] Ets, AP-1, C/EBP, NF-IL-6* (c) CCAAT, PU.1, NF-IL-6, AML1 (h) Tartrate-resistant acid phosphatase (TRAP) Iron response region* (h, m) [205] Gardner-Rasheed feline sarcoma (fgr) PMA response region* (h, m) [206] Secretory leukocyte protease inhibitor (SLPI) [207] C/EBP ETS (h) [26] Leukotriene B4 receptor [208] Genes primarily expressed in granulocytes Myeloperoxidase MyNF1 (PEBP2/CBF)* (m) [36, 37, 209] Sp1-thyroid hormone-retinoic acid* (h) Neutrophil elastase TATA, CCAAT (h, m); PU.1 (h) [27, 37, 126] MyNF1,* C/EBP,* PU.1,* c-myb* (m) G-CSF receptor PU.1,* C/EBP * (h) [28] Migration inhibitory factor-related proteins TATA, CCAAT, Sp1 (h) [210] (MRP8-MRP14) Proteinase 3 (PR3) TATA, Myb, C/EBP, PU.1* (h) [ ] Azurocidin (AZU) TATA, Myb, C/EBP, PU.1* (h) [211, 212] Cathepsin G PU.1 (h) [214] Myeloid zinc finger protein (MZF1) Retinoic acid [215] Genes expressed at a high level in myeloid cells Inducible nitric oxide synthase (inos) LPS-induced NF- B,* interferon regulatory factor 1 [167, 168, 173, 216] (IRF-1),* TATA, NF-IL-6, hypoxia regulatory element (HRE),* Stat1 * (m) Aminopeptidase N (CD13) Myb,* Ets-1* TATA, PU.1, Sp1, AP-2 (h) [217, 218] Urokinase type plasminogen activator (upa) PU box* (m) [ ] PU.1, AP-1, AP-2 (h) Alpha-1-antitrypsin LPS induced NF- B* (h) [222] Monokine induced by IFN- (mig) IFN- response element-1 ( RE-1)* (m) [223, 224] PU.1 Ets,* AP-1,* Oct* (m) [85] 5-Lipoxygenase-activating protein (FLAP) AP-2 (h) [225, 226] 5-Lipoxygenase (5-LO) [227] *, Functional site; h, human; m, mouse; c, chicken. 154 Journal of Leukocyte Biology Volume 63, February 1998

3 TABLE 2. Transcription Factor Binding Sequences in Myeloid Genes Promoter Sequence References Sequences that bind C/EBP h M-CSF R TTTCCAAA [17] h GM-CSF R TTCCCAAT [29] h G-CSF R GTCGCAAT [28] m elastase TGGGCAAT [27] Consensus TTNCCAAT or or or or GGNGCAAA Sequences that bind PEBP2/CBF h, m, M-CSF R TGTGGTTGC [17, 56] m, h GM-CSF TGTGGTCAC [57] m elastase TGTGGCCAG [37] m myeloperoxidase TGTGGTTAG [36, 37] h CD36 TGTGGTTTG [58] Consensus TGTGGTTNG or or or TGTGGCCNC Sequences that bind PU.1 h M-CSF R AAGGGGAAGAAGA [92] h GM-CSF R ATGAGGAAGCAGG [29] h G-CSF R CAAGAGAAGTTCC [28] TTCAGGAACTTCT h elastase GAGAGGAAGTGGA [27, 126] h CD11b AAGGAGAAGTAGG [110] h MSR AAAGAGAAGTGAA [98] hfc RI AAGAGGAAGGAAA [113, 114] hfc RIII AAGAGGAAGGAAA [120] h c-fps/fes AGGAGGAAGCGCG [194, 195] h IL-1 AAGGGGAAAAGAG [228] Consensus h, Human; m, mouse. NRRRRGAAGNRNR or or C or or NTCRRGAAANTNC andthe enhancer core (G)TGGT/AT/AT/A(G), led to the independent isolation of C/EBP, the defining member of this leucine zipper transcription factor family [13, 14]. Important properties of the family are a conserved cysteine that allows disulfide cross-linking between homo- and heterodimers [15] the potential to interact with other proteins [16 18], and as shown in the discovery of the family, a certain degree of binding promiscuity, at least in vitro [19]. Although the C/EBP family was originally thought to be restricted to hepatocytes and adipocytes, the cloning of other family members, particularly NF-IL-6 (C/ EBP ) and its chicken homologue NF-M, showed diverse tissue activity [15, 20 23]. In the hematopoietic system C/EBP has only been found in the myeloid lineage and may be important in the early commitment of progenitor cells because the factor is down-regulated when myelomonocytic cells are induced to differentiate [24, 25]. A myeloid-specific family member, C/EBP, which regulates gene expression in promyelocytes, has recently been cloned [26]. C/EBP binding sites (Tables 1 3) are critical for myeloid-specific expression of neutrophil elastase [27] and the receptors for macrophage colonystimulating factor (M-CSF), granulocyte colony-stimulating factor (G-CSF), and GM-CSF [17, 28, 29]. NF-IL-6 (C/EBP ) was first described as an IL-1-inducible factor necessary for in vitro expression of IL-6 and the protein avidly binds the SV40 enhancer core and CCAAT sequence as well as the IL-6 promoter sequence, AGATTGCACAATCT [21], again demonstrating a range of in vitro recognition sequences for members of this family. A second gene, NF-IL-6, with high homology to NF-IL-6 [15, 30, 31], forms heterodimers with NF-IL-6 in vitro and synergizes with NF-IL-6 when cotransfected. Recombinant NF-IL-6 binds cis regions in several proteins expressed in myeloid cells (Table 1) and although the transcription factor is not expressed, or expressed at a very low level in normal mouse tissues, it is rapidly induced by tissue injury, infection, or inflammatory cytokines such as IL-1 and TNF- [21, 22]. It is also up-regulated during induced macrophage differentiation of cell lines and cultured primary human monocytes [25]. C/EBP and C/EBP have been disrupted in mice by homologous recombination. In the case of C/EBP [32], the overwhelming phenotype is the severe defect in glycogen synthesis, which results in death within 8 h after birth. In these newborn mice the monocytes and macrophages develop normally but there are no mature neutrophils and very low levels of G-CSF receptor [33] consistent with the demonstrated in vitro role of C/EBP in expression of this receptor [28]. The C/EBP null mice [34, 35] develop a phenotype similar to that of mice overexpressing IL-6, and the high IL-6 levels in the uninduced null mice indicate that NF-IL-6 is not necessary for IL-6 activity in vivo, and may even act as an inhibitor. The mice have large lymph nodes, caused by a net increase in B cells, defects in lymphoid and myeloid proliferation, and become susceptible to Listeria, Salmonella, and Candida infections. However, macrophage cell numbers, phagocytic ability, production of TNF-, IL-1, interferon- (IFN- ), IL-10, inducible nitric oxide synthase (inos), nitric oxide (NO), Nramp, Fc R, MHC II, Mac I, macrophage inflammatory protein-1 (MIP-1 ), and GM-CSF are all within normal limits. G-CSF production is impaired but administering the growth factor to the infected animals does not help. Antibody responses appear to prevail over cell-mediated immunity and a Th2 type response is mounted against Candida albicans. Splenic macrophages fail to release NO 2 and IL-12 is undetectable in the serum of null mice after Candida infection, indicating that macrophage function is severely compromised. The Runt/PEBP2/CBF family A sequence (Table 2) necessary for expression of murine myeloperoxidase binds two proteins, designated Myeloid Nuclear Factor 1 (MyNF1, ) found in murine myeloblast, but not erythroid, B, or fibroblast cell lines [36, 37]. The proteins are members of the polyoma virus enhancer binding factor 2/core binding factor (PEBP2/CBF) group, the mammalian homologues of the Drosophila runt domain gene family [38 40]. Members of the family bind as heterodimers in which only the subunit binds DNA and the subunit increases the binding affinity 5- to 10-fold [38, 39]. Three related subunit loci have been identified in both humans and mice [41 43], whereas the subunit is encoded by a unique gene [44]. A human homologue of PEBP2 B/CBF 2 is the AML1 gene Clarke and Gordon Myeloid-specific gene expression 155

4 TABLE 3. Proximal Promoter Binding Factors in Myeloid-Restricted Genes Gene Site Factor References h M-CSF R 40 to 50 PU.1 [17, 56, 92] 66 to 72 AML1 (PEBP2) 76 to 84 C/EBP h MSR 50 to 67 AP-1/Ets [97, 98] 185 to 198 PU.1 h CD14 94 to 100 PU.1? [12, 104] 104 to 110 Sp1 h CD11b 16 to 22 PU.1 [110, 111] 58 to 64 Sp1 hfc RI 1 to 13 PU.1 [113, 114] 13 to 23 (C/EBP?) mfc RIII 40 to 50 PU.1 [120] 78 to 88 MyE? h GM-CSF R 41 to 53 PU.1 unidentified [29] 54 to 70 C/EBP m myeloperoxidase 288 to 292 MyNF1, (PEBP2) [36, 37] m neutrophil elastase 50 to 45 Myb [27, 37] 60 to 53 C/EBP 70 to 66 PEBP2 79 to 69 PU.1 h G-CSF R 36 to 39 PU.1 [28] 43 to 46 PU.1 49 to 45 C/EBP involved in translocations reported in 30 40% of acute myeloid and B cell leukemia [45, 46], suggesting a role for this family member in hematopoiesis. Alternatively spliced forms of the subunit, which may act as inhibitors, and a disrupted subunit, have also been implicated in myeloid leukemia [47 49]. Although runt domain proteins can be found in most human and mouse tissues, PEBP2 B/CBF 2/AML1 is expressed at a high level in hematopoietic tissues such as thymus and spleen [43, 48, 50]. In adult mouse bone marrow runt-like factors are confined to myeloid and B cells; expression is absent in erythroid cells [51]. Embryos with a disrupted PEBP2 B/ CBF 2 gene appear normal to about day 11.5 but die as the major site of hematopoiesis shifts from the yolk sac to the fetal liver at around day 12.5 [52, 53]. There is a complete absence of fetal liver hematopoiesis and progenitor cells cannot be cultured from the yolk sacs or the livers of mutant embryos. Disruption of the subunit results in the same lethal phenotype [54, 55]. Because MyNF1 clones have not yet been reported, it is not possible to know whether or not the proteins are equivalent to already cloned murine and human Runt/PEBP2/CBF genes or are myeloid-specific members of the family. Evidence for a possible myeloid-restricted factor emerged when a Runt/PEBP2/ CBF family protein, which binds a sequence necessary for monocyte-specific expression of the M-CSF receptor promoter, could be isolated from a monocyte cell line, but not from T, B, or epithelial lines [56]. Also a myeloid-specific isoform can be detected in Western blots of mouse bone marrow subpopulations [51]. In addition to myeloperoxidase [36, 37] and the M-CSF receptor [56], PEBP2/CBF regulates neutrophil elastase [37], both human and mouse GM-CSF [57], and CD36 [58] (Tables 1 3). The ETS family Ets-1, the cellular counterpart of a region in the avian E26 retrovirus [59, 60] is the founding member of the Ets (E twenty-six specific) family, which now numbers about thirty related proteins [61]. Recent structural studies reveal that the factors belong to the winged helix-turn-helix (whth) group [62 64]. Near the carboxyl terminus, all Ets-related factors share a highly conserved DNA binding region of 85 highly basic amino acids, termed the ETS domain [65, 66]. The most divergent ETS domain is in PU.1, with only 35% amino acids identical to those of Ets-1. The transcriptional activation domains are in non-conserved, amino-terminal regions, implying an ability to interact with diverse accessory factors or components of the transcription machinery. The ETS domain binds DNA as a monomer, which appears to be sufficient for in vitro DNA binding to a purine-rich region with a central GGA core [61, 66, 67]. That the sequences which flank the central GGA element can dramatically affect binding and confer specificity was shown when, in gel mobility retardation assays, PU.1 bound the sequence GCAGGAAGT and Ets-1 bound the sequence AGAGGAACT, but not vice versa [68]. PU.1 binding sequences identified in myeloid-restricted genes are listed in Table 2. A purine-rich region, the PU box [69], in the mutated enhancer of a lymphoid-specific SV40 variant was used as a probe to clone the 272-amino acid transcription factor, PU.1 [67]. Unlike other Ets family proteins, the PU.1 protein contains a central PEST region (proline, glutamic acid, serine, threonine) associated with protein instability [70]. In cell lines, PU.1 is confined to myeloid leukemia and B cells; it is not expressed in T cells [67, 69]. When normal mouse bone marrow is examined, myeloid and erythroid precursors, macrophages, 156 Journal of Leukocyte Biology Volume 63, February 1998

5 and megakaryocytes are positive for PU.1, whereas no PU.1 is detected in mature granulocytes, osteocytes, or vascular endothelium [71]. Freshly isolated CD34 human bone marrow cells, as well as mature human peripheral blood monocytes have PU.1 message but, unlike the negative mouse bone marrow neutrophils [71], human peripheral blood neutrophils express high levels of PU.1 [72]. The presence of PU.1 in the early stages of B cell, erythroid, and myeloid hematopoietic lineages is suggestive of a role for the transcription factor in the regulation of hematopoiesis. In vitro, expression of PU.1 in human and mouse cells is coincident with myeloid differentiation and interference with PU.1 binding reduces myeloid and erythroid precursors [72 74]. In mice, targeted disruption of the ETS domain of PU.1 results in multiple hematopoietic abnormalities [75 77]. A homozygous mutation is either lethal for the embryos at day 18.5 [75] or results in death from septicemia 2 days after birth [76]. At birth mice lack mature lymphoid and myeloid cells and osteoclasts; however, erythrocytes and megakaryocytes are present. If kept alive with antibiotics, low numbers of T cells and neutrophils develop, but mature B cells and macrophages are not detected. Macrophages cannot be cultured from fetal or neonatal spleens or livers. Lack of osteoclasts results in the osteopetrotic phenotype, which can be rescued by marrow transplantation after which macrophage and osteoclast numbers are restored to normal in transplanted mice, which can survive for 6 months [77]. It is interesting to note that, in the PU.1-deficient mice described in these studies, except for defects in hematopoiesis and osteoclastogenesis, development appears normal. Animals, despite the lack of macrophages, can be born alive and survive for 2 weeks, suggesting that macrophages are not critical for embryonic tissue remodeling or non-hematopoietic cell growth and differentiation. Finally, if the phenotype of the PU.1 knockout can be taken as an indication of the existence of unidentified myeloid/lymphoid precursor, there have been few reports of such multipotential cells. Two laboratories report the isolation of a cell type, one purified from mouse fetal liver [78], and one purified from mouse bone marrow cells [79], that differentiate into mixed B cell and macrophage colonies. In some Ets family DNA binding studies, members interact with other factors and in some cases cooperation appears to be required for activity [61, 68, 80]. For example, in B cells NF-EM5/PIP (nuclear factor enhancer motif 5/PU.1 interaction partner), which forms a complex with PU.1 and is required for PU.1 activity may be restricted to B cells, for the protein is not expressed in macrophage cell lines [81, 82]. As yet there is no significant evidence of a requirement for interaction between PU.1 and another factor in myeloid cells. In one study the PU.1 recognition sequence of the human GM-CSF receptor appears to bind PU.1 complexed with another unidentified protein [29] and in another study, when NF-IL-6 was isolated after PU.1 was used as a probe to clone PU.1 interacting proteins, no physical interaction was detected between the two, although there was synergistic transactivation [83]. An intriguing in vitro study [84] suggests the possibility that PU.1 may bind to the TFIID protein of the RNA polymerase II basal transcription complex. Removal of the 75-residue amino-terminal activation domain of PU.1 severely reduces the binding. MYELOID-RESTRICTED GENES Listed in Table 1 are cloned genes, each with a proportion of cell-type expression restricted to the myeloid lineage and the genes (and one virus) are divided into four groups according to the reported pattern of the restricted expression. Of course, none of the groupings is absolute and none is meant to indicate that any particular gene has never been found expressed in non-myeloid cells. The groupings, and the reported studies, are divided thus as a convenient way to attempt an understanding of the transcriptional distinction between the myeloid cell types and between myeloid and non-myeloid cell types. The groups are as follows: genes primarily expressed in the macrophage lineage; primarily expressed in macrophages and granulocytes; primarily granulocyte restricted; and those which have a myeloid-specific transcript or regulation but are expressed in other cell types. A good example of the latter group is the transcription factor PU.1 in which the control of expression in myeloid and B cells has been examined at the transcription level [85]. The regulation of the transcription of about half of the genes listed has been studied, some, at least for the proximal promoter region, in detail. The findings from 10 of these will be summarized in order to derive some overall concept of the control of myeloid-specific expression. Three, M-CSF receptor, MSR, and CD14 are from the macrophage group; four, CD11b, Fc RI, Fc RIII, and GM-CSF receptor are from the macrophage/granulocyte group; and three, myeloperoxidase, neutrophil elastase, and G-CSF receptor are from the granulocyte group. Genes primarily expressed in macrophages The M-CSF receptor is required for the differentiation, proliferation, and survival of mononuclear cells [86] and expression is restricted to monocytes/macrophages [86, 87] and placental trophoblast cells [88]. Cell-type-specific expression is achieved through the use of different promoters [89, 90]. No expression is found in granulocytes [87]. Mouse and human genes lack a TATA box [91] and a 416-bp fragment ( 416 to 126) of the human proximal promoter directs macrophage-specific reporter gene expression; reduction to 85 bp retains 82% of activity [92]. A PU.1 binding sequence ( 50 to 40, Table 2) is essential for full monocyte-specific activity of these promoters [91, 92]. A second essential sequence (Table 2) only 20 bp upstream from the PU.1 site binds AML1, and separated by only 3 bp, C/EBP [17, 56]. Although both the PU.1 and the AML1 binding regions are necessary for full macrophagespecific expression, there is no evidence of interaction between the two factors. The regions are almost exactly conserved in the mouse proximal promoter [17]. However, there may be interaction between C/EBP and AML1 because, although mutations of the binding sites of either factor reduces promoter activity by 85%, and a double mutation has the same effect, simply increasing the distance between the two sites from 3 to 5 bp or to 10 bp reduces proximal promoter activity by 80% [17]. In Clarke and Gordon Myeloid-specific gene expression 157

6 addition, neither C/EBP nor the heterodimeric AML1 alone is able to transactivate the 416-bp promoter-reporter construct in a non-myeloid cell line, all three proteins are required [17]. The macrophage scavenger receptor (MSR) is expressed in mature macrophages, is developmentally activated, and can be up-regulated by growth factors and cytokines [93, 94]. The human [94, 95] and mouse [96] genes lack a TATA box. As expected, AP-1 binding regions ( 4.5 kb, 4.1 kb, and proximally, at 67 to 50), each in this case with adjacent Ets-like motifs, are involved in TPA induction of the receptor during differentiation [97]. It is interesting that these TPA response regions appear to be specific, for the sequences are not able to confer TPA induction on promoters other than MSR; thus the AP-1 sites may be involved in both up-regulation and tissue-restricted expression. A proximal promoter region ( 245 to 49), which contains a PU.1 site ( 198 to 181, Table 2) and a TPA response region ( 67 to 50) restricts reporter gene expression to macrophages, and surprisingly, HeLa cells [97, 98]. Mutation of either the PU.1 region or the Ets-like sequence in the TPA response region reduces myeloid-specific transfection efficiency by 50%, indicating the importance of both purine-rich regions for expression. Mutation of the AP-1 site, as predicted, eliminates the ability of the proximal promoter to increase activity in response to TPA, but has little effect on macrophage-specific expression. However, a double mutation of the PU.1 region and the AP-1 site reduces promoter activity by 85%. The presence of the ubiquitous AP-1 binding proteins in HeLa cells appears adequate to transactivate the MSR proximal promoter for a mutated AP-1 site abolishes the expression in HeLa cells. CD14 binds the Gram-negative lipopolysaccharide (LPS) complexed with LPS binding protein [99] as well as a unique CD14 binding constituent produced by Gram-positive bacteria [100]. The binding is a potent activator of myeloid cells and can induce the release of immunoregulatory and inflammatory mediators such as TNF-, IL-1, IL-6, and prostaglandins [101]. A soluble and a GPI-anchored membrane form of CD14 have been described [102]. The soluble form appears to be involved in the response of epithelial and endothelial cells and the membrane-bound form in the response of myeloid cells [103]. Whether or not the two forms are expressed in specific cell types is not yet known. Neither the mouse nor the human genes contain a TATA box and the 5 region of both genes are highly homologous [12, 104]. In the mouse, myeloid cellspecific expression is retained in 257 bp ( 257 to 61) in humans, as few as 128 bp of the 5 region ( 128 to 80) confer macrophage-specific transient expression. Sp1 binds the human CD14 promoter region at 110 and its binding sequence (GGGCTGG) is critical for expression, since mutation of the site eliminates Sp1 binding and reduces transient transfection efficiency by 70% [12]. A putative PU.1 binding site is only 5 bp proximal to the functional Sp1 site; whether or not the PU.1-like site is functional has yet to be determined [12]. Genes primarily expressed in both macrophages and granulocytes CD11b (CR3), the subunit of the heterodimeric CD11b/CD18 (Mac1) integrin cell surface receptor [105], is expressed in monocytes, macrophages, neutrophils, natural killer cells [106], and a subset of B cells [107]. Immature precursors do not express CD11b and very little is expressed by cell lines such as HL60 and U937, unless induced to differentiate [108]. The promoter of the human gene lacks a TATA box and 412 bp of 5 sequence ( 412 to 83) is sufficient to direct myeloid-specific reporter gene expression [109]. Indeed, with only 92 bp of promoter region there is still significant myeloid-restricted expression. Both the PU.1 site ( 22 to 16, Table 2) and the Sp1 site ( 64 to 58) are necessary for full expression [110, 111] for mutations in the binding regions of either significantly reduces but does not eliminate transient expression of the proximal promoter. In addition, Sp1 binds the CD11b promoter in a myeloid cell line and not in HeLa cells, suggesting that a ubiquitous factor is able to contribute to myeloid specificity [111]. Fc RI (CD64) is a high-affinity receptor for monomeric mouse IgG2a and human IgG1. A soluble form of the human receptor has also been characterized [112, 113]. A 39-bp sequence ( 168 to 132) numbered from the start of translation, which is not promoter- or cell-type restricted, is necessary and sufficient to confer IFN- inducibility. The macrophage specificity of Fc RI induction seems to be a property of the core promoter for a 23-bp region that confers specificity on the thymidine kinase minimal promoter overlaps the major transcription start of the human gene [ ], which like so many myeloid-restricted genes does not have a TATA box. The myeloid-specific region is conserved in the mouse proximal promoter, has a PU.1 site (Table 2), which after mutation results in a 70% loss in activity, and the PU.1 binding region appears to bind a second unidentified protein [ ]. Fc RIII (CD16) is a low-affinity receptor for Ig, with two isoforms, Fc RIII-A and FcR III-B, encoded by two distinct genes in humans. Only one gene, Fc RIII-A, is found in the mouse [116]. In humans, the transmembrane A gene is expressed in activated monocytes, macrophages, and natural killer cells, but not neutrophils, whereas the GPI-anchored III-B gene, with greater than 95% sequence identity in coding and flanking regions, is uniquely expressed in neutrophils [117, 118]. The single murine gene is expressed in the three cell types, thus an analysis of the transcription control of this Fc receptor provides a unique opportunity to determine any control sequences that distinguish neutrophil and macrophage celltype expression. A cis region ( 943 to 850) important for natural killer cell expression has been identified in the human gene [119]. There are no TATA boxes and six or more transcription start sites in each of the three genes. In the single murine gene a 51-bp sequence ( 90 to 39) confers macrophage-specific expression on a core promoter [120]. Within this myeloid-restricted region, mutations in either the PU.1 site ( 50 to 40, Table 2) or the MyE (myeloid E box) site ( 88 to 78) reduce transfection efficiency by 85 90%; a double mutation does not reduce the efficiency further. The proteins that bind the myeloid E box have yet to be identified. The 1800 bp before the start of translation of the two human genes are almost identical, with only 38 substitutions, a block of 8 nucleotides deleted from the A gene, and a triplet deletion from the B gene; 26 of the substitutions are within 500 bp of the ATG 158 Journal of Leukocyte Biology Volume 63, February 1998

7 [117]. For both genes, the proximal 200 bp retain transient differential natural killer and granulocyte expression, and within these two promoter regions there are only 10 nucleotide changes, 8 of which introduce new consensus binding sites for transcription factors. Consequently, the same putative binding sites have a different distribution in the two proximal regions (Table 1). The chain of the receptor for GM-CSF is expressed in mature normal human neutrophils, macrophages, eosinophils, and in peripheral blood cells of patients with myeloid leukemia [121]. A soluble form of the human chain has been described [122]. The human gene lacks a TATA box and has a proximal promoter region ( 70 to 49) that retains myeloid specificity and binds PU.1 ( 53 to 41, Table 2), as well as C/EBP ( 70 to 54, Table 2) immediately upstream [29]. There is no evidence of interaction between the two, for mutation of both has no additional effect beyond the 70 80% reduction in efficiency observed after single mutations. In addition, either recombinant PU.1 or C/EBP is able to transactivate the proximal promoter; both together do not increase efficiency. Genes primarily expressed in granulocytes Myeloperoxidase is a microbicidal enzyme present in the primary granules [123]. The human [124] and mouse [125] genes lack TATA boxes. A murine promoter region ( 302 to 281, Table 2), necessary for activation only in myeloid cell lines, binds at least two myeloid nuclear proteins, one of which, MyNF1, is a member of the PEBP2/CBF family and possibly a myeloid-specific member of this family [36, 37]. The human and mouse myeloperoxidase promoter regions are highly homologous [36], making it likely that transcription control of both will be similar. Neutrophil elastase is a serine protease expressed in developing granulocytes in the bone marrow and stored in azurophilic granules. In the human gene, 106 bp of 5 region are sufficient to confer myeloid-specific expression and a 30-bp sequence, which binds PU.1, is required for this specificity [126]. The mouse gene has an equally short, highly homologous myeloidspecific proximal promoter that binds ( 70 to 66, Table 2) the PEBP2/CBF family member MyNF1 [37]. Functional sites for PU.1 ( 79 to 69, Table 2), C/EBP ( 60 to 53, Table 2), and Myb ( 50 to 45) are also present [27]. In contrast to all the myeloid genes so far described, the human and mouse neutrophil elastase genes have TATA boxes. The G-CSF receptor is expressed in neutrophils and, like the M-CSF receptor, also in the placenta [ ]. Unlike the M-CSF receptor both human cell types use the same G-CSF receptor major transcription start [28]. The mouse and human genes do not have TATA boxes and have a highly homologous 200-bp proximal promoter, which in the human gene contains a functional C/EBP site ( 49 to 45, Table 2) and two functional PU.1 sites ( 36 to 39, 43 to 46, Table 2) [28]. CIS REQUIREMENTS FOR MYELOID-SPECIFIC GENE EXPRESSION In making a summary of the information derived from these published studies, it is important to remember that the starting point for most of the investigations has been the minimum cis region required for myeloid-specific transient expression in cell lines, and this in all cases meant the 5 proximal promoter region, which was usually less than 300 bp from the start of transcription. Although some studies of negative regulation have been reported [94, , 176] in general, positive regulation has been investigated and the expression and binding data are from in vitro assays. The importance of particular sites was usually ascertained by mutating the site in promoter fragments of the order of 0.5 to 1 kb in length. The transactivation experiments, such as activation of the M-CSF receptor or CD11b promoter by PU.1 in HeLa cells [17, 110], were all done by co-transfection of the recombinant transcription factor and the relevant promoter region fused to a reporter gene. These transient transfection experiments are probably not indicative of any effects on the regulation of the endogenous gene from the introduced expression of a transcription factor, and indeed none were reported. Nor are they necessarily a prediction of stable expression patterns, whether in cell lines or transgenic mice. Given the concentration on the minimal requirements for transient transfection, it appears that in most cases the proximal promoter regions that confer myeloid-specific expression have some or all of the following characteristics. (1) There is no TATA box and multiple transcription start sites. (2) There is a functional PU.1 site close to a major transcriptional start site consistent with a possible role in the initiation of transcription. (3) The PU.1 site is required for transient transfection of reporter constructs, although abolition of PU.1 binding does not completely eliminate reporter gene activity. (4) Recombinant PU.1 is able to transactivate promoter constructs in nonmyeloid cell types. (5) There is binding of another factor(s) close to the PU.1 binding region and this second region is important for transient transfection. The second factor(s) is able to transiently transactivate the proximal promoter in nonmyeloid cells. (6) There is rarely evidence of interaction between PU.1 and the nearby binding factors, although both are needed for full transient activity of the promoter constructs. (7) Myeloid-specific transient expression is conferred on a heterologous promoter by a region of about 50 bp, which contains a binding site for PU.1 and at least one other factor. The proximal promoter binding factors that were shown to be critical for myeloid-restricted expression of the genes described in this section, and the position of the binding sites relative to the major transcription start site, are listed in Table 3. As can be seen, in addition to PU.1, the binding proteins are usually C/EBP, PEBP2/CBF, and Sp1. MATTERS FOR CONSIDERATION Several points arise from the studies described above, which will probably be clarified as more experimental work is reported. The first and most obvious, is the importance of the factor binding sites for the expression of genes in vivo, leading to the possibility of incorporating the crucial regions into a minimal promoter construct, which can be used to target the myeloid lineage. Some of the promoter fragments described have been tested in transgenic mice. The in vivo importance of Clarke and Gordon Myeloid-specific gene expression 159

8 the PU.1 binding site in the MSR promoter was clearly demonstrated when the 296-bp proximal promoter region, which confers myeloid-specific transient expression, was ligated to the 400-bp ( 4.5 to 4.1) TPA response region and used to target a reporter gene in mice [133]. A mutation in the PU.1 site abolishes expression of the reporter gene in four lines of transgenic mice, and reduces expression in two others to a low level [133]. Without the PU.1 mutation, the reporter gene is expressed in the macrophage lineage in five of seven lines. Two CD11b proximal promoter regions have been tested in vivo [134]. The 92-bp proximal promoter region, which retains myeloid-specific transient expression in cell lines and contains PU.1 and Sp1 sites, when used to produce four lines of transgenic mice, results in transgene expression in the bonemarrow-derived granulocytes and monocytes of only one line, and this line also expresses the transgene in lymphocytes. In addition, a 300-bp CD11b promoter region, which directs high myeloid-specific expression in cell lines, is able to direct expression in only one of four founder lines, this time, not in bone marrow cells, but in the thymus. Clearly, these CD11b minimal regions require other regulatory sequences in addition to the PU.1 and Sp1 sites, for appropriate expression and/or to avoid position effects. Although the myeloid-specific transcription factor binding sites in the promoter of the human gp91- phox gene have not yet been described, the 450-bp region, which is adequate to produce an IFN- response in cell lines [135], directs reporter transgene expression in the bone marrow monocytes of 11/15 lines. Reducing the promoter fragment to 138 bp eliminates monocyte expression [136], indicating the presence of critical regions between 450 and 138 bp. Another obvious question that arises when cotransfection with PU.1 is shown to transiently transactivate a PU.1- dependent promoter in a cell line that does not normally express PU.1 is whether or not B cell lines, which have endogenous PU.1, will be able to transiently express the reporter construct. In an attempt to answer this question [94, 98] it was shown that 700 bp of the human scavenger receptor 5 region are adequate to direct expression in both mouse and human macrophage cell lines, but that there is no expression in a B cell line, although the B cell line could be shown to express functional PU.1. The possibility that it is repression that may prevent activation of the scavenger receptor promoter in B cells was suggested when a mutated MSR promoter that increases transient transfection activity in macrophage cell lines, presumably by eliminating the binding of a repressor, allowed expression in a B cell line. And finally, is it possible to distinguish between the regulatory regions of genes expressed in macrophages and granulocytes or both? Given the results so far reported, the answer is no. As can be seen in Tables 2 and 3, aside from PU.1, which is likely to be required for all myeloid-specific expression, the families of binding proteins are distributed throughout these myeloid-restricted genes, without any distinction between lineages. In addition, there were few attempts to make such a distinction in reports. It might be possible to identify critical lineage-restricted cis elements if regulatory regions were transfected into myeloid progenitor cells and expression were monitored as the cells were induced to differentiate. Information about the lineage differences could also be obtained from further analyses of the human Fc RIII-A and B genes, since it has been shown in transgenic mice that about 6 kb of the 5 region of the A gene, and about 19 kb of the B gene, target the macrophage and neutrophil lineages, respectively [118]. More precisely, there is the intriguing possibility of sorting out the granulocyte and macrophage-specific pattern of transcription factors, which might confer cell-type-specific expression in the 200 bp of the human Fc RIII-A and B gene promoters, since there are only 10 nucleotide differences between the two promoter regions [117]. The pieces of the model that will eventually describe myeloid-specific gene expression are being discovered at an increasing rate and significant contributions will probably be made from studies of regions beyond the proximal promoter, and from studies of all regions in transgenic mice. MYELOID-SPECIFIC TRANSGENIC MICE Several characteristics of the reported myeloid-specific transgenic mice listed in Table 4 should be noted. First, there is no pan-myeloid or macrophage transgene, that is, no gene that directs expression in myelomonocytic precursors, or in monocytes, which continues to be expressed as the cells differentiate and migrate to all tissues. The heterogeneity of the macrophage population, and the fact that cloned genes have tended to be selected on the basis of their involvement in immunological processes, has biased the myeloid promoter studies toward TABLE 4. Transgene Myeloid-Specific Transgenic Mice and One Rabbit* References Mo MuLV LTR_m GM-CSF [ ] Mo MuLV LTR_m GM-CSF _diphtheria toxin [140] Mo MuLV LTR_h CD14 genomic clone [141] Visna virus LTR_CAT [142] Visna virus LTR_h ApoE [143] h Cathepsin G genomic clone [144] h Cathepsin G genomic clone _PML-RAR [145] h MRP8 genomic clone _h bcl2 [146] h MRP8 genomic clone _PML-RAR [147] h c-fps/fes genomic clone [148] c Lysozyme genomic clone [132, ] c Lysozyme genomic clone, anti-hel Ig double transgenic [153, 154] c Lysozyme genomic clone _h 15 lipoxygenase* [155, 156] hfc RI genomic clone [157] hfc RIIIA genomic clone [118] hfc RIIIB genomic clone [118] h Scavenger receptor promoter _h GH [133] h M-CSF receptor promoter _GAP-C [158] h M-CSF receptor promoter _ETS [158] h gp91-phox promoter _h GH [136, 160] h gp91-phox promoter _SV40 large T antigen [136, 160] h CD11b promoter _h CD4 [134] h CD11b promoter _ -galactosidase [161] h CD11b promoter _h GH [161] h CD11b promoter _PML-RAR [162] h Lysozyme promoter _m IFN- receptor chain [163] h Lysozyme promoter _CAT [164] 160 Journal of Leukocyte Biology Volume 63, February 1998

9 some highly responsive genes that are induced and/or amplified in recruited or resident myeloid subpopulations. Second, most of the transgenes studied, which perhaps by chance are almost all controlled by human promoters, have at least one report of aberrant expression, either because the transgene does not duplicate the endogenous human or mouse gene expression pattern or because the transgene is expressed in an unexpected cell type. However, it should also be noted that this is not necessarily critical, because if a regulatory region can be shown to reliably direct high level transgene expression in any particular subpopulation of myeloid cells, or to direct expression under any particular set of environmental stimuli, the regulatory region can be used to study a biological problem. In fact, considering the comprehensive range of organs and tissues that contain macrophages, a pan macrophage transgene may be undesirable, or at least irrelevant, for the study of a particular biological problem, for example, the role of macrophages in the thymus. Third, the analysis of the transgene expression patterns is almost always incomplete because the nature of the ubiquitous and active macrophage would require an exhaustive number of analyses of isolated cell types and tissue sections, in both control and immunologically challenged animals, to identify the complete array of macrophage populations able to express any particular transgene. Viral promoters The earliest reports of transgene expression in myeloid cells use the Moloney murine leukemia virus long-terminal repeat (LTR) to overexpress murine GM-CSF in order to study the long-term effects of overexposure to this growth factor [ ]. Surprisingly, expression of the transgene is restricted to three types of tissue macrophages generated by local proliferation. The transgenic mice have elevated levels of macrophages in the pleural and peritoneal cavities, in muscle tissues, and in the developing eye, which leads to blindness. These locally proliferating macrophages express high levels of TNF-, IL-1, and bfgf, and many of the mice die from muscle wasting at 2 4 months. The fact that the MoMuLV LTR directs expression to the macrophage-like cell of the eye, the hyalocyte, implicated in the tissue remodeling of the post-natal mouse eye was later [140] used to ablate this particular subpopulation of cells, and show that the hyalocyte is vital both for the induction of cell death, as well as the clearing of tissues that would normally be transient in the developing eye. In this case a polycistronic construct consisting of the MoMuLV LTR, with the coding region of the diphtheria toxin A chain inserted into the 3 region of the GM-CSF gene was used to direct ablation. The MoMuLV LTR was again used [141] after failure to get expression in four transgenic mouse lines that contained the human CD14 gene with about 4.2 kb of the upstream region. Ligation of the MoMuLV LTR to the 5 end of the same CD14 region produced a pattern of expression unlike that of the MoMuLV LTR GM- CSF construct, but which nevertheless provides evidence of the importance of CD14 as the primary mediator of endotoxic shock. The combination of CD14 regulatory regions and MoMuLV LTR directs expression in monocytes, peritoneal macrophages, neutrophils, and T cells; B cells do not express the transgene. A second viral control region used to target myeloid cells was that of the visna virus [142], a lentivirus of sheep that can infect many cells of the monocyte/macrophage lineage, although viral gene expression is induced only in mature macrophages. The pattern of reporter gene expression in whole tissues indicates a lack of expression in resident cells and induction of expression in some populations of activated macrophages. As in sheep, expression is detected in the brain but the transgene is aberrantly expressed in splenic lymphocytes. However, in a study designed to determine the contribution of macrophageexpressed apoe to decreased atherosclerosis, the 622-bp LTR, when used to direct expression of the complete human apolipoprotein E gene in apolipoprotein E-null mice, results in transgene expression in macrophage-derived foam cells in atherosclerotic lesions without detectable transgene expression in lymphocytes [143]. From these reports it can be concluded that the MoMuLV LTR and the visna virus LTR can be useful promoters/ enhancers for use in myeloid-directed expression, although the expression pattern may depend on the particular construct used. Genomic clones The next group of transgenes discussed will be the seven, cathepsin G, MRP8, c-fps/fes, lysozyme, Fc RI, Fc RIII-A, and Fc RIII-B in which the transgenes were the entire genomic clones, with both 5 and 3 genomic sequences. Because of this, it is not possible to know which regions are responsible for the observed patterns of expression. Nevertheless, the reports do provide some insights into the requirements for myeloidspecific expression. Two of the genomic clone transgenes, human cathepsin G and human MRP8 were expected to target the granulocyte lineage. The 6-kb human cathepsin G genomic clone, with about 2.5 kb of 5 and 0.8 kb of 3 sequence, appropriately targets the fetal liver and early myeloid cells in adult bone marrow [144]. Insertion of the PML-RAR fusion gene into the 5 untranslated region plus a mutation in the cathepsin G translation start produced a transgene used to determine the level of expression and latent period required for the onset of acute promyelocytic leukemia [145]. The second genomic clone transgene, human MRP8, was used to examine the role of the bcl-2 gene in the control of the equilibrium that exists in the rapidly turning over neutrophil population. Transgenic mice were generated with a polycistronic transgene in which the bcl-2 cdna was inserted into the 3 untranslated region of the human MRP8 gene, which had about 1.5 kb of 5 and 1.5 kb of 3 MRP8 sequence [146]. The regulatory elements necessary for correct neutrophil expression appear to be present in the construct because the transgene was expressed only in early myeloid cells and neutrophils in bone marrow, blood, and spleen. Use of a similar transgene with a PML-RAR fusion insert provided an additional mouse model of promyelocytic leukemia [147]. In summary, the constructs controlled by human cathepsin G and human MRP8 regulatory regions successfully target the granulocyte lineage and, although in both cases the entire genomic clone was used, the most likely regulatory regions are only 2.5 kb or less from the start of Clarke and Gordon Myeloid-specific gene expression 161