Myeloid expression of cytochrome P450 4F3 is determined by a lineage-specific alternative promoter.

The cytochrome P450 4F3 (CYP4F3) gene encodes two functionally distinct enzymes that differ only by the selection of exon 4 (CYP4F3A) or exon 3 (CYP4F3B). CYP4F3A inactivates leukotriene B4, a reaction that has significance for controlling inflammation. CYP4F3B converts arachidonic acid to 20-hydroxyeicosatetraenoic acid, a potent activator of protein kinase C. We have previously shown that mRNAs coding for CYP4F3A and CYP4F3B are generated from distinct transcription start sites in neutrophils and liver. We therefore investigated mechanisms that regulate the cell-specific expression of these two isoforms. Initially, we analyzed the distribution of CYP4F3 in human leukocytes and determined a lineage-specific pattern of isoform expression. CYP4F3A is expressed in myeloid cells and is coordinate with myeloid differentiation markers such as CD11b and myeloperoxidase during development in the bone marrow. In contrast, CYP4F3B expression is restricted to a small population of CD3+ T lymphocytes. We identified distinct transcriptional features in myeloid, lymphoid, and hepatic cells that indicate the presence of multiple promoters in the CYP4F3 gene. The hepatic promoter depends on a cluster of hepatocyte nuclear factor sites 123-155 bp upstream of the initiator ATG codon. The myeloid promoter spans 400 bp in a region 468-872 bp upstream of the ATG codon; it is associated with clusters of CACCT sites and can be activated by ZEB-2, a factor primarily characterized as a transcriptional repressor in cells that include lymphocytes. ZEB-2 interacts with C-terminal binding protein and Smads, and this would provide opportunities for integrating environmental signals in myelopoiesis and inflammation.

The ability of cells to attain differentiated stages in development and to alter their functional phenotype in inflammatory settings depends on lineage-specific and cytokine-dependent transcription factors that either activate or repress target genes. Distinct patterns of transcription factors are associated with the myeloid and lymphoid lineages in hematopoiesis (1) and with the induction of pro-inflammatory genes in host defense (2). Transcription factor activity is context-dependent; a repressor of one gene can activate a different gene associated with an opposing function or alternative cell lineage (3). This has implications for the coordinated induction and repression of sets of genes that are functionally related. ZEB-1 (␦EF1, Zfhx1a) and ZEB-2 (SIP1, Zfhx1b) are recently characterized two-handed zinc finger transcription factors (4,5) that may function as both gene silencers and activators. ZEBs repress transcription of genes that include IL-2, CD4, and GATA-3 in lymphocytes (6 -8), whereas their role in myeloid cells has not been determined. The C-terminal binding protein (CtBP) 1 is a corepressor for ZEBs and other factors but does not bind DNA directly (9). Recent evidence suggests that ZEBs can activate the vitamin D 3 receptor gene in colon carcinoma cells (10).
Cytochrome P450 4F3 (CYP4F3) functions to control inflammation by inactivating leukotriene B 4 (LTB 4 ) with high specificity (11)(12)(13), but the mechanisms that regulate CYP4F3 gene expression are not known. LTB 4 is a potent chemoattractant of myeloid cells (14)(15)(16)(17) and has been implicated in the pathogenesis of multiple inflammatory diseases (18)(19)(20)(21). LTB 4 is synthesized from arachidonic acid by the sequential action of 5-lipoxygenase and leukotriene A 4 hydrolase, and the chemoattractant activity of LTB 4 is mediated by a high affinity G protein-coupled receptor designated BLT1 (22). The major pathway for the catabolism and inactivation of LTB 4 in human neutrophils involves microsomal -hydroxylation by CYP4F3 (11). This reaction generates 20-OH LTB 4 , which can further inhibit LTB 4 activities by down-regulating BLT1 (23). The ability of LTB 4 to amplify an inflammatory response is, therefore, counterbalanced by the expression of CYP4F3.
Recently we cloned the CYP4F3 gene and demonstrated that alternative splicing generates isoforms that differ in functional properties and tissue distribution (24,25). Selection of exon 4 generates the neutrophil isoform (CYP4F3A), which has a low K m for LTB 4 of Ͻ1 M. Selection of exon 3 instead of exon 4 generates an alternative isoform (CYP4F3B), which has a 44fold lower efficiency of inactivating LTB 4 . CYP4F3B is expressed in liver and has a preference for arachidonic acid as a substrate, which it converts to . This has significance because 20-HETE is a potent bioactive mediator in certain tissues; it activates protein kinase C and has roles in regulating cell proliferation, vascular tone, and natriuresis (26). These opposing capacities of the CYP4F3 isoforms to generate an active mediator  or inactivate one (LTB 4 ) allow for versatility of function but demand strict con-trols. CYP4F3 transcription and alternative splicing must be regulated to ensure that the appropriate isoform is generated in the correct setting.
To understand the regulation of CYP4F3 expression, we examined its distribution in maturing populations of human bone marrow cells and identified the splicing pathways. CYP4F3A is expressed in myeloid cells; its expression is coordinate with known myeloid differentiation markers such as CD11b and also increases concomitantly with myeloperoxidase during development. In contrast, CYP4F3B is expressed in lymphocytes, and expression is restricted to a small population (ϳ10%) of CD3ϩ T cells. We determined that alternative promoters regulate lineage-specific expression of CYP4F3 isoforms in myeloid cells, lymphoid cells, and liver. Surprisingly, activity of the myeloid-specific promoter could not be accounted for by known myeloid transcription factors including PU.1 and MZF-1, but it could be activated by ZEB-2 and CtBP. The results suggest new roles for these proteins in myeloid transcription. ZEBs and CtBP are regulated by activated Smads and NADϩ, respectively, and therefore have the capacity to integrate various environmental signals during development or inflammation.

EXPERIMENTAL PROCEDURES
Cells-Human bone marrow was obtained from discarded filters used in the processing of normal donor marrow from transplantation at the Massachusetts General Hospital. Peripheral blood samples were collected from healthy donors according to established guidelines. Approval for the use of human samples was obtained from the Institutional Review Board of the Massachusetts General Hospital/Partners. Granulocytes and lymphocytes were separated by Ficoll-Hypaque (Amersham Biosciences) density gradient centrifugation. HL60 cells (human promyelocytic cell line) were maintained in RPMI containing 10% fetal bovine serum. The cells were treated with 1.3% Me 2 SO for 4 days to induce granulocytic differentiation before transfection experiments. HepG2 cells (human hepatoma-derived cell line) and COS 7 cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum.
Flow Cytometry-Total bone marrow mononuclear cells and peripheral blood granulocytes and lymphocytes were stained with allophycocyanin, percyphycoerythrin, or fluoroisothiocyanate-conjugated control IgGs or monoclonal antibodies directed to CD34, CD38, CD33, CD13, CD16, CD15, CD11b, CD14, CD8, and CD3 (BD Biosciences) to define hematopoietic cell subsets in a double or tri-color labeling. Surfacelabeled samples were then fixed in 0.1% formaldehyde and permeabilized with 0.1% Triton X-100 to perform intracytoplasmic staining with anti-CYP4F3 antibody. Production, purification, and characterization of polyclonal rabbit antibodies against the C-terminal domain (amino acids 410 -520) of CYP4F3, which recognize both isoforms (CYP4F3A and CYP4F3B), has been described previously (24). Intracellular staining was performed on unlabeled or surface-labeled cells after fixation and permeabilization. Samples were incubated with 5 g/ml anti-CYP4F3 polyclonal antibody or 3 g/ml control rabbit IgG for 30 min at room temperature and were then incubated with 1 g/ml goat antirabbit phycoerythrin-conjugated or fluoroisothiocyanate-conjugated polyclonal antibodies (Sigma). Tri-or four-color-labeled samples were then analyzed by flow cytometry using the FACScalibur instrument and the Cellquest program (BD Biosciences). The data were plotted as single-parameter histograms or bivariate dot plots with logarithmic amplification.
5Ј-Rapid Amplification of cDNA Ends (5Ј-RACE) Analysis-The 5Јuntranslated region (5Ј-UTR) sequences of CYP4F3 transcripts in different RNA samples were determined by 5Ј-RACE (Invitrogen). Firststrand cDNA synthesis was performed with primer 1 (5Ј-CCCCAACC-AGCAGCAGGAGCAGCC-3Ј, ϩ79 to ϩ56 relative to adenosine of first methionine). First-round PCR was with the Anchor Primer and primer 2 (5Ј-ATGCTGCCATTGGCCAAA-3Ј, ϩ48 to ϩ31). Second-round PCR was with the Universal Amplification Primer and primer 3 (5Ј-GACA-GGCTCAGCTGTGGC-3Ј, ϩ20 to ϩ3). The PCR conditions were 94°C for 1 min, 50°C for 1 min, and 72°C for 1 min; 30 cycles were followed by 1 cycle with a 10-min extension time. PCR products were subcloned using a TOPO TA cloning kit (Invitrogen) and sequenced. A minimum of 10 RACE clones was analyzed from each RNA sample.
Construction of Luciferase Reporter Plasmids and Site-directed Mutagenesis-CYP4F3 promoter-luciferase reporter constructs were made using the pGL3 basic vector (Promega). Defined segments of the 5Јflanking region of the CYP4F3 gene were amplified by PCR. KpnI and NheI sites were added to the 5Ј-ends of sense and antisense primers, respectively, and PCR products were ligated into the corresponding sites of pGL3 and sequenced. Site-directed mutagenesis was performed using the GeneEditor System (Promega) according to the manufacturer's instructions.
Transient Transfections and Luciferase Assays-HL60 cells were treated with 1.3% Me 2 SO for 4 days and electroporated with a Gene Pulser (Bio-Rad) at settings of 290 V and 960 microfarads. Each sample (2.3 ϫ 10 7 cells/ml in complete growth medium plus 1.3% Me 2 SO, 300 l/cuvette) was electroporated with 30 g of pGL3 luciferase reporter plasmid, and 10 g of ␤-galactosidase expression plasmid (pcDNA3.1/Myc-His-LacZ, Invitrogen) was included as a control for transfection efficiency. HepG2 cells and COS 7 cells were grown to 50 -60% confluency in 6-well plates and transfected using SuperFect Reagent (Qiagen). HepG2 cells were transfected with 1 g of luciferase plasmid plus 1 g of ␤-galactosidase plasmid/well. COS 7 cells were transfected with a total of 2.5 g of DNA/well (0.5 g of luciferase plasmid, 0.5 g of ␤-galactosidase, 0.5 g of CtBP expression vector, and 1 g of ZEB expression vector); the total DNA added was made up to 2.5 g with PCDNA3 plasmid (Invitrogen) in control experiments without expression vectors. HepG2 and COS 7 cell lysates were prepared in reporter lysis buffer (Promega) 48 h after transfection. HL60 cell lysates were prepared 14 -18 h after the addition of DNA. Luciferase activity was measured with a luciferase assay system (Promega) and a luminometer (Monolight 2010, Analytical Luminescence Laboratory) and normalized to the ␤-galactosidase activity in each sample as determined by a ␤-galactosidase enzyme assay system (Promega). The results represent the mean of at least three independent experiments. Expression vectors for human ZEB-1 and ZEB-2 (in pCS2 with an N-terminal Myc epitope tag) were generous gifts from Dr. A. Postigo and Dr. D. Dean (5). An expression vector for human CtBP-1 was a generous gift from Dr. R. Goodman (37).
Electrophoretic Mobility Shift Assays-The wild type (WT) probe corresponds to a 38-bp region between positions Ϫ812 and Ϫ775 of the CYP4F3 promoter (5Ј-CACTCAGTAGGTGCTCTTTAAGAGCAGG-TGTCACACAG-3Ј) and includes the putative bipartite ZEB binding site. The mutant probe has each AGGTG sequence converted to CCCGA (5Ј-CACTCAGTCCCGACTCTTTAAGAGCCCCGATCACACA-G-3Ј). Sense and antisense oligonucleotides were annealed, and doublestranded oligonucleotides were then labeled with [␥-32 P]ATP and T4 polynucleotide kinase (Promega). Nuclear extracts were prepared from COS 7 cells transfected with Myc-tagged ZEB-2 using the nuclear extract kit from ActiveMotif. The DNA binding assay (24-l final volume) was carried out at 4°C using a Gelshift Kit from ActiveMotif with buffers A1 and B1. The reaction contained 5 g of nuclear extract and 100 pg of 32 P-labeled double-stranded oligonucleotide (10 5 cpm). For supershift experiments, a mouse monoclonal antibody to c-Myc (Santa Cruz Biotechnology) was included in the DNA binding assay. A mouse monoclonal antibody to lamin AϩB 2 (Zymed Laboratories Inc.) was used as a negative control. The reaction mixtures were separated on a 4% polyacrylamide gel prepared in 0.25% TBE buffer, and the gel was then dried and exposed to Kodak X-Omat AR-5 film for 10 h.

CYP4F3 Expression in Human Bone
Marrow-To understand the regulation of the CYP4F3 gene and its isoforms, we initially determined the expression of the CYP4F3 protein in different populations of developing hematopoietic cells. Non-CYP4F3 in Hematopoietic Cells erythroid nucleated bone marrow cells were isolated from four individual donors and stained with affinity-purified fluorescent-labeled anti-CYP4F3 for analysis by flow cytometry. Analysis of total non-erythroid bone marrow cells indicated that 40 -60% expressed CYP4F3 protein (Fig. 1A), with 30 -40% exhibiting high and 10 -20% exhibiting intermediate fluorescence. To identify cell populations for further analysis, bone marrow cells were next broadly separated into the subdivisions R1, R2, and R3 ( Fig. 1B) based on the properties of size (forward scatter) and granularity (side scatter). These populations were analyzed for CYP4F3 expression using rabbit IgG as a negative control (Fig. 1C). A small but consistent population of cells (3%) demonstrated positive staining for CYP4F3 in R1 (pink), the region of low forward scatter, and low side scatter that includes lymphocytes and early progenitors. The region R2 (green) includes a mixed population of cycling progenitors, monocytes, and maturing myeloid cells, and 64% of these cells were positive for CYP4F3. This region was further separated into the subdivisions R4, R5, and R6 (Fig. 1D), and a progressive increase in CYP4F3 expression (49, 66, and 76% positivity, respectively) was observed with increasing granularity (maturity). Cells with the highest levels of CYP4F3 expression were concentrated in R3 (red), a region of low forward scatter and high side scatter that is composed primarily of mature granulocytes (Fig. 2C). Essentially all of these cells (99%) were pos-itive. The results indicate that expression of CYP4F3 is concomitant with myeloid maturation.
We next compared the expression of CYP4F3 in bone marrow to that of known myeloid proteins using a panel of antibodies against surface markers for hematopoietic differentiation and maturation in conjunction with anti-CYP4F3 (Fig. 2). A representative experiment is shown in Fig. 2A. A high percentage of total bone marrow cells are double-positive for CYP4F3 and myeloid markers (for example, 48% CD33ϩ, 37% CD11bϩ, and 32% CD15ϩ cells express CYP4F3). Less than 2% of cells are double-positive for CYP4F3 and lymphoid markers such as CD2, CD3, and CD8. Interestingly, this population of lymphocytes expressing CYP4F3 was consistently observed, and it is apparent that some of the cells have a high level of expression. The data presented for total bone marrow cells in Fig 2C; cells immunolabeled for specific myeloid and lymphoid surface marker antigens were analyzed to determine the percentage that express CYP4F3. An average of 92% CD15ϩ, 80% CD11bϩ, 20% CD34ϩ, 15% CD8ϩ, and 10% CD3ϩ cells were positive for CYP4F3. Overall, the results confirm that CYP4F3 is expressed primarily in myeloid cells but indicate that it is also expressed in a small population of CD34ϩ progenitors and CD3ϩ T cells.
To determine whether CYP4F3 is differentially expressed in maturing myeloid cells, we compared CYP4F3 and MPO expression in CD33ϩ gated populations (Fig. 3). Cells that were negative or low for MPO also had low expression of CYP4F3, and CYP4F3 expression increased in parallel with MPO. This is consistent with the increases in CYP4F3 fluorescence ob-served as a function of cell granularity (Fig. 1) and confirms that CYP4F3 expression is strongly associated with myeloid differentiation.
Expression of CYP4F3 in Peripheral Blood Leukocytes-Flow cytometry was used to compare the expression level of CYP4F3 in different cell types in peripheral blood (Fig. 4). Granulocytes exhibit high fluorescence (Fig. 4A), whereas lymphocytes exhibit low fluorescence similar to negative control staining with IgG (Fig. 4B). When peripheral leukocytes were analyzed for myeloid and lymphoid markers that included CD15, CD14, CD3, and CD8, the results paralleled those in bone marrow in that essentially 100% of CD15ϩ and CD14ϩ cells but only ϳ10% of CD3ϩ and CD8ϩ cells co-expressed CYP4F3 (not shown). A HL60 promyelocytic leukemia cell line exhibited intermediate fluorescence (Fig. 4C). Analysis of blood from two patients with eosinophilia gave identical results; CYP4F3 was expressed at highest levels in the eosinophils followed by neutrophils and then monocytes (Fig. 4, D and E). Eosinophilia samples were analyzed for CD16, CD45, and CD49D in addition to the markers listed above (not shown).
Altenative Promoters Regulate Tissue-specific Expression of CYP4F3-We have previously shown that functionally distinct splice forms of CYP4F3 exhibit a tissue-specific distribution, with CYP4F3A expressed in neutrophils, and CYP4F3B expressed in liver (24,25). We suggested that the 5Ј-UTR of CYP4F3 transcripts is also tissue-specific and might indicate the use of alternative promoters in different cell types (24). To further investigate this possibility, RNA samples from selected cell populations were analyzed by 5Ј-RACE to determine the sequence of the 5Ј-UTR and to identify the transcription start sites (Fig. 5A). Similar results were obtained in peripheral blood granulocytes, bone marrow myeloid cells isolated at early (BM-1, CD11bϩ CD14Ϫ) or later (BM-2, CD11bϩ CD14ϩ) stages of development, and also HL60 cells; a single transcription initiation site was identified 519 bp upstream of the ATG initiation codon (start site A). The 5Ј-UTR requires splicing to assume its location upstream of the ATG codon in mature myeloid transcripts, and the only variation observed relates to selection of the 5Ј-splice donor site. Two alternative splice junctions are used in bone marrow myeloid cells, apparently with similar efficiency, to generate 5Ј-UTRs of 40 or 34 bp. The 40-bp 5Ј-UTR was the only form observed in peripheral blood granulocytes. In HepG2 cells, transcription is initiated at a site located 71 bp upstream of the initiation codon (start site B), representative of the liver phenotype (24). The relative positions of transcription start sites A and B are shown schematically in Fig. 5B.
We investigated whether the use of alternative start sites A and B in myeloid cells and liver was representative of the use of alternative tissue-specific promoters. To test this hypothesis 5Ј-flanking regions of the CYP4F3 gene were cloned into a luciferase reporter vector and transfected into HL60 cells differentiated with Me 2 SO or HepG2 cells (Fig. 5C). In HL60 cells, maximum luciferase expression was obtained using a promoter region that extends 350 bp upstream of start site A (construct A6). A different pattern was observed in HepG2 cells; the 350-bp region upstream of start site A has low activity (construct A6), whereas the region between start sites A and B has much higher activity (construct B3). Maximum luciferase expression in HepG2 cells was obtained using a promoter region that extends 630 bp upstream of start site B (construct B6). This construct includes start site A and overlaps the myeloid promoter, but it has low activity in HL60 cells, probably because splicing of exon 1 (the 5Ј-UTR) is required for luciferase expression in these cells. The splicing process may be impaired due to modification of the 3Ј-splice junction in promoter-luciferase constructs. Complete deletion of the 3Ј-splice junction abolishes activity in HL60 cells but has no significant effect in HepG2 cells (compare construct B6 with B4).
Constructs A1-A8 were designed to eliminate a requirement for splicing after transcription initiation from start site A; exon 1 of CYP4F3 (the myeloid 5Ј-UTR) is linked directly to the ATG initiation codon of luciferase. Initially, the two 5Ј-splice donor sites in exon 1 were excluded (constructs A1-A4). Subsequently, it was determined that the region containing these splice sites (position Ϫ500 to Ϫ468) was required for full transcriptional activity in HL60 cells (constructs A5-A8). In contrast, the 5Ј-UTR downstream of start site B is not required for activity in HepG2 cells (compare construct B3 to B5).
The Myeloid Promoter Is Activated by ZEB-2 and CtBP-We initially sought to identify a role for known myeloid transcription factors in the regulation of CYP4F3 gene expression. Potential binding sites were predicted by inspection of the DNA sequence of the myeloid promoter (region Ϫ872 to Ϫ468) and are summarized in Table I. These include putative sites for the factors MZF-1 (based on the core sequence GGGGA, positions Ϫ602, Ϫ538, and Ϫ515)) and PU.1 (permissible variations of the GAGGAA core sequence, positions Ϫ575 and Ϫ518). Surprisingly, mutations to these sites did not reduce activity (Table I). We then focused on the region in the 5Ј-UTR between Ϫ500 and Ϫ468, which is an active element of the myeloid promoter. This region contains a putative ZEB target site; ZEB-1 (␦EF1) and ZEB-2 (SIP1) are two-handed zinc finger transcription factors that bind to bipartite CACCT(G) sequences with a variable orientation and spacing of nucleotides (4). Two inverted CACCT sites (AGGTG) with a spacing of 11 bp are located between positions Ϫ488 and Ϫ468, a location that exactly matches the two splice junctions (Fig. 6A). Furthermore, a similar bipartite motif is observed further upstream between positions Ϫ804 and Ϫ783, a location that is within a region of the promoter required for activity in HL60 cells (compare constructs A5 and A6 in Fig. 5C).
The effect of ZEB expression on CYP4F3 promoter activity was measured in COS 7 cells (Fig. 6B). The cells were cotransfected with a ZEB expression vector and construct A6, which contains the myeloid promoter and both bipartite ZEB sites at positions Ϫ488 and Ϫ804. A 12-fold induction of activity was observed if an expression vector for CtBP was included. No induction of luciferase activity was observed in the absence of CtBP. The response was dependent on the amount of ZEB expression vector used for transfection and decreased in an approximately linear fashion from 1 to 0.25 g vector. No significant differences were observed in the response to expression vectors for ZEB-1 or ZEB-2. Mutations to the ZEB sites were investigated using constructs A5 and A3, which contain one bipartite site at position Ϫ488, or Ϫ804, respectively. ZEB and CtBP induce a 6-fold increase in activity of construct A5, and this is abolished by deleting the region Ϫ500 to Ϫ468 (mut1). ZEB and CtBP induce a 15-fold increase in activity of construct A3, and this can be reduced to 2-fold by a site-specific mutation that converts AGGTG to CCCTG at position Ϫ804 (mut2). Construct A3 does not contain any splice junctions, and the effect of mut2 on transcription is, therefore, clearly separable from any effects on splicing. The activity of these mutants in HL60 cells is shown in Table I. Deletion of the ZEB site in exon 1 (⌬Ϫ500 to Ϫ468) results in a 71% decrease in activity of construct A5, and the site-specific mutation at position Ϫ804 results in an 88% decrease in activity of construct A3.
The expression of ZEB-1 and ZEB-2 in myeloid cells was analyzed by isoform-specific reverse transcription-PCR (Fig.  7A). Bone marrow CD11bϩ cells, peripheral blood granulocytes, and HL60 cells express ZEB-2 but not ZEB-1. HepG2 cells and CD3ϩ T cells preferentially express ZEB-1, and CtBP was detected in all samples. ZEB-1 and ZEB-2 are highly related and bind to identical DNA sequences, but they may have different transcriptional activity due to differences in their strength of interaction with other proteins (5). A gel retardation assay was then used to demonstrate binding of Myc-tagged ZEB-2 to the CYP4F3 promoter (Fig. 7B). A 32 Plabeled oligonucleotide probe corresponding to the region between Ϫ812 and Ϫ775 of the promoter forms a binding complex with ZEB-2 (Fig. 7B, lane 2 2 and 4), and may be accounted for by an interaction of ZEB-2 with other cellular proteins.
The distribution of the two isoforms of CYP4F3 containing exon 4 (4F3A) or exon 3 (4F3B) was examined by reverse transcription-PCR using the same RNA samples that were assayed for ZEB-1 and ZEB-2 (Fig. 7A). We previously showed that 4F3A is selectively expressed in peripheral blood granulocytes, whereas 4F3B is selectively expressed in liver (24). All myeloid cell populations analyzed in this study expressed 4F3A and not 4F3B. This includes bone marrow cells isolated at early (BM-1, CD11bϩ CD14Ϫ) or later (BM-2, CD11bϩ CD14ϩ) stages of development and HL60 cells. Conversely, the only isoform detected in CD3ϩ T cells was 4F3B.
Transcription in Non-myeloid Cells-A cluster of sites essential for transcription in HepG2 cells was identified in a region of the hepatic promoter 45-83 bp upstream of start site B (Table I). These include a site for the hepatic factor HNF4 (position Ϫ155) and two sites for the HNF3/FREAC family of factors (positions Ϫ137 and Ϫ123), which are active in a variety of tissues including liver and kidney (27). We previously reported (24) the presence of two TATA-like sequences at positions Ϫ549 (upstream of myeloid start site A) and Ϫ105 (upstream of hepatic start site B). However, mutations at these sites have low or minimal effect on the activity of promoter constructs (Table I), and the alternative promoters in the CYP4F3 gene may, therefore, belong to a class that does not depend on a TATA box. The hepatic and myeloid promoters overlap between positions Ϫ710 and Ϫ468, and we identified a single-bp mutation at position Ϫ632, which selectively reduces transcription in HepG2 cells. The mutation disrupts a consensus site for AP4 (CAGCTG). The location of transcription factor sites shown to be important for CYP4F3 expression in HL60 and HepG2 cells are shown in Fig. 8A.
Peripheral blood CD3ϩ T lymphocytes were isolated to high purity (Ͼ99%) by flow cytometry, and 5Ј-RACE was performed on RNA preparations to determine the transcription start site of CYP4F3 in these cells. The sequences of 10 5Ј-RACE clones were identical and identify a 19-bp 5Ј-UTR with the sequence AATCTAAGTGCTTACCAGG. This indicates a novel transcription start site (designated start site C) located 219 bp upstream of the ATG initiation codon, intermediate between start sites A and B (Fig. 8A). It is separated from the myeloid transcription start site A by 300 bp, raising the possibility that distinct promoter elements are recruited to enable expression in lymphoid cells. This was not characterized in detail. Transcription initiation from start site C requires splicing to link the 5Ј-UTR to the ATG initiation codon and is independent of transcription initiation from start site B.
The splicing pathways of CYP4F3 in myeloid cells, liver, and T cells are shown schematically in Fig. 8B. Selection of exon 4 is the critical event that determines CYP4F3 function in myeloid cells. Recently we cloned the CYP4F3 gene (24) and showed that a unique feature of exon 4 is a 25-bp repeat that duplicates FIG. 6. A, a bipartite ZEB site in exon 1 (position Ϫ488) is composed of two inverted CACCT sequences that correspond with the location of 5Ј-splice donor sites. A bipartite ZEB site at position Ϫ804 is composed of one CACCT and one CACCTG sequence (both inverted). The ATG translation initiation codon is located at the beginning of exon 2. B, regulation of CYP4F3 promoter activity in COS 7 cells. Control cells were transfected with a promoter luciferase construct (A6, A5, or A3). Activation of the promoter by ZEB was measured by cotransfecting cells with an expression vector for ZEB-1, ZEB-2, or CtBP-1, and luciferase activity was recorded as a-fold induction relative to the control (error bars indicate S.D., n ϭ 3). Cells were transfected with 1 g of ZEB expression vector per well unless otherwise indicated. Construct A5 was mutated by deleting bp Ϫ500 to Ϫ468 (mut1). Construct A3 was mutated by converting the sequence AGGTG (position Ϫ804) to CCCTG (mut2).

FIG. 7.
A, distribution of ZEB and CYP4F3 isoforms. RNA was isolated from bone marrow CD11bϩ CD14Ϫ cells (BM-1), bone marrow CD11bϩ CD14ϩ cells (BM-2), peripheral blood granulocytes (PBG), HL60 cells, HepG2 cells, and peripheral blood CD3ϩ T cells isolated to high purity (Ͼ99%) by flow cytometry. The RNA was analyzed by reverse transcription-PCR using primers specific for ZEB-1, ZEB-2, CYP4F3A, and CYP4F3B or primers that recognize both isoforms of CtBP (CtBP-1 and CtBP-2). The size of the PCR product is indicated. B, electrophoretic mobility shift assay for binding of ZEB-2 to promoter DNA. the 5Ј-splice donor sequence (Fig. 8C). This is shown here for comparison with the myeloid 5Ј-UTR (exon 1), which contains two alternative 5Ј-splice junctions. The sequence at the 5Јsplice junction of exon 4 is an inverted TACCTG, and this is duplicated to create a sequence resembling a bipartite ZEB binding motif. For example, ZEB site selection experiments demonstrate that nucleotide usage at positions 2-4 of the CAC-CT(G) consensus is essentially 100%, but that the C at position 1 can be replaced by T in 14% of cases (28). It is, therefore, possible that similar factors bind to these unique elements in the myeloid 5Ј-UTR and exon 4, a feature that could enable coupling of myeloid transcription and splicing.

DISCUSSION
The CYP4F3 enzyme functions to inactivate nanomolar concentrations of LTB 4 for chemotaxis by catalyzing its -hydroxylation (11)(12)(13). 20-OH LTB 4 binds to BLTR1 with the same high affinity as LTB 4 and inhibits responses to LTB 4 by downregulating the receptor (23). Determining the distribution of CYP4F3 expression should, therefore, identify cells that participate in the modulation of LTB 4 function. An understanding of mechanisms that regulate CYP4F3 gene expression might elucidate pathways that control LTB 4 -mediated inflammation. We used flow cytometry to analyze the distribution of CYP4F3 in human bone marrow cells and observed an expression pattern that was closely correlated with developing myeloid cell populations (Figs. 1-3). Increases in CYP4F3 fluorescence were concomitant with increased cell granularity and acquisition of myeloid differentiation markers such as myeloperoxidase. In peripheral blood, CYP4F3 is expressed at highest levels in eosinophils followed by neutrophils and then monocytes (Fig. 4).
The presence of CYP4F3 in eosinophils and its strong association with myeloid differentiation in bone marrow have not previously been shown but are consistent with its functional involvement with LTB 4 regulation. LTB 4 is a chemoattractant for human eosinophils (17), and eosinophils isolated from patients with hypereosinophilic syndromes express abundant cytosolic phospholipase A 2 , 5-lipoxygenase, and BLT1 (29). Interestingly, an average of 20% CD34ϩ progenitor cells in the bone marrow exhibit CYP4F3 reactivity (Fig. 2C), and this may be due to populations of early myeloid-committed cells that have been identified in previous studies by coexpression of CD34 and CD11b (30). An early onset of CYP4F3 expression in bone marrow would be consistent with a role in maintaining low basal levels of LTB 4 in addition to regulating LTB 4 activity during inflammation. CYP4F3 may protect against an inappropriate amplification of LTB 4 production as developing myeloid cells express BLT1 and LTB 4 biosynthetic enzymes.
We recently cloned the CYP4F3 gene and postulated that alternative promoters regulate transcription in liver and neutrophils because distinct 5Ј-UTRs are generated in each location (24). Many cytochrome P450s including CYP4F3 are expressed in liver, whereas expression in myeloid cells is unusual. The existence of an alternative promoter for myeloid expression is now confirmed using HL60 cells as a model system. Overlapping but distinct promoter regions were identified in HL-60 and HepG2 cells transfected with luciferase reporter constructs containing 5Ј-flanking regions of the CYP4F3 gene (Fig. 5). The myeloid promoter identified in HL60 cells extends over a relatively short region (350 bp) upstream of transcription start site A (the myeloid start site), requires exon 1 (50 bp immediately downstream of the start site) for full activity, and a TATA box is not essential. Similar features have been observed in promoters for other myeloid-specific genes (31). The presence of alternative promoters represents another unusual feature of the CYP4F3 gene and is comparable with the situation described for CYP19 (aromatase) expression that has been intensively investigated (32). Transcription of CYP19 is regulated by different promoters in different tissues and results in transcripts with different 5Ј-UTRs, although the protein encoded by these transcripts is always the same.
The myeloid promoter of CYP4F3 includes potential sites for traditional myeloid transcription factors such as MZF-1 and PU.1, but mutations to these sites did not reduce activity in differentiated HL60 cells. An active region of exon 1 contains a bipartite ZEB site consisting of two inverted CACCT sequences separated by 11 bp (Fig. 6A), and deletion of this region reduces transcriptional activity by 71% (Table I). Interestingly, the location of these inverted CACCT sequences corresponds to the location of two 5Ј-splice donor sites in the pre-mRNA. A similar bipartite ZEB site is located at position Ϫ804, 285 bp upstream of the transcription start site, in another active region of the promoter. Mutagenesis of the site at Ϫ804 reduces transcriptional activity by 88% (Table I), and this effect is clearly independent of any potential interference with splicing. Two related ZEB genes have been identified, and both are expressed in hematopoietic cells; ZEB-1 (␦EF1) is present in T cells, and ZEB-2 (SIP1) is present in B cells (5). Previous studies have focused on the role of ZEB genes in lymphocytes and have not specifically investigated myeloid cells. We used isoform-specific PCR to determine that ZEB-2, but not ZEB-1, is expressed in peripheral blood granulocytes and developing myeloid cells in the bone marrow (Fig. 7A). This is the first indication that ZEB-2 has the capacity to act as a myeloid transcription factor. ZEB proteins have been characterized primarily as transcriptional repressors that bind to the corepressor CtBP (33). However, ZEB activates vitamin D 3 receptor (VDR) gene transcription in colon carcinoma cells, and this activity was confirmed in COS 7 cells transfected with a ZEB expression vector and the VDR promoter (10). We utilized this approach to demonstrate activation of the CYP4F3 myeloid promoter by ZEB and observed a 10 -15-fold induction of activity with either ZEB-1 or ZEB-2 if a CtBP expression vector was cotransfected (Fig. 6B). This induction was dependent on the amount of ZEB expression vector used for transfection and could be inhibited by mutagenesis or deletion of the ZEB sites. Activation of the VDR promoter did not require CtBP (10), suggesting that ZEB activity is more sensitive to context in the CYP4F3 promoter. Gel retardation assays demonstrate that ZEB-2 forms a binding complex with CYP4F3 promoter DNA, which is dependent on intact CACCT sites (Fig. 7B). The assays were performed with nuclear extracts from COS 7 cells that were transfected with Myc-tagged ZEB-2. Cotransfection of these cells with CtBP was not required to demonstrate ZEB-2 binding to DNA. The function of CtBP, therefore, seems to be unrelated to DNA binding and more likely involves modification of interactions with other proteins in the promoter complex. The capacity of ZEBs to interact with multiple proteins presents opportunities for CYP4F3 gene regulation. ZEB-2 was originally identified as a Smad-interacting protein and designated SIP1 (34). An inverted Smad3 site (GCCCAGAC) is located at position Ϫ629 between the two bipartite ZEB sites in the CYP4F3 promoter, and Smads are activated by transforming growth factor-␤, which is an active cytokine in myeloid development and inflammation (35,36).
Recently it was proposed that CtBP can function as a redox sensor for transcription by detecting changes in nuclear NAD ϩ / NADH ratios (37). ZEB-CtBP interactions and transcriptional activity could be enhanced with agents such as CoCl 2 , which perturb cellular oxidation. Other reports agree that CtBP is regulated by NAD but suggest that NAD ϩ and NADH are equally effective (38). Total NAD ϩ and NADH levels may vary in response to cellular energy status or DNA damage, and this may lead to changes in CtBP activity. Both mechanisms of CtBP regulation have the potential to be engaged during inflammation. It is well documented that CtBP acts as a corepressor of a number of transcription factors in addition to ZEBs (9). Our results suggest that CtBP acts as a coactivator in specific settings. Interestingly, Drosophila CtBP can act as a transcriptional corepressor or coactivator in a context-dependent manner (39). It is possible that ZEB/CtBP represses transcription from the myeloid promoter of CYP4F3 in certain cell types such as lymphocytes or that it represses transcription of other genes in myeloid cells.
A small population of CD3ϩ T cells (ϳ10%) expresses CYP4F3 in bone marrow and peripheral blood, and this appeared to contradict the conclusion that CYP4F3 expression is restricted to hematopoietic cells of myeloid lineage by a myeloid-specific promoter. However, 5Ј-RACE analysis identified a novel transcription start site in these cells located at position Ϫ219 (designated start site C). A distinct lymphoid promoter region may, therefore, be involved. We have previously described two isoforms of CYP4F3 that are generated by alternative splicing (25). Selection of exon 4 generates CYP4F3A, which has a low K m for LTB 4 of Ͻ1 M. Selection of exon 3 generates CYP4F3B, which has a preference for arachidonic acid as a substrate, which it converts to 20-HETE. Isoformspecific PCR demonstrated that CYP4F3A is the isoform ex-pressed in myeloid cell populations detected by anti-CYP4F3 in flow cytometry experiments (Fig. 7A). In contrast, the isoform detected in CD3ϩ T cells was CYP4F3B. LTB 4 is functionally active on some lymphocyte populations (40,41), but generation of 20-HETE by CYP4F3B in T cells may be of greater importance than inactivation of LTB 4 . The ability of 20-HETE to activate MAP kinase pathways and stimulate cell proliferation (42,43) could have implications with respect to T cell activation or oncogenesis.
Selection of exon 4 is the critical event that determines CYP4F3 function in myeloid cells (25). CYP4F2 is the only other member of the human CYP4F family to contain homologues of both exons 3 and 4 (other characterized members contain a homologue of exon 3 only), but the latter does not appear to be used as a substrate for splicing. This is unexpected because the CYP4F2 and CYP4F3 genes are highly related. A unique feature of CYP4F3 is a 25-bp repeat that duplicates the 5Ј-splice donor sequence of exon 4 (Fig. 8C), and this is similar to exon 1 (the myeloid 5Ј-UTR), where two alternative 5Ј-splice donor sequences correspond with the location of a bipartite ZEB site. The ZEB transcription factors have a number of unusual features; they can act as transcriptional repressors or activators (6,7,8,10); they interact with multiple proteins including CtBP and Smads (33,34); they bind with two zinc finger regions to two target sites (4); they can act over long distances and may serve as bridging molecules (44). It is interesting to speculate that they might also play a role in splicing. There is growing evidence that transcription and splicing are coupled in many genes (45), and this would have particular relevance to myeloid CYP4F3 expression and its regulation in oxidative stress and inflammation.
Expression of the CYP4F3 gene involves alternative choices at the level of transcription initiation, transcription termination, and splicing. Alternative splicing of CYP4F3 represents the first recorded example of functionally different enzymes being encoded within a single CYP gene (25); substrate specificity is determined by selection of exon 4 (CYP4F3A) or exon 3 (CYP4F3B). 3Ј-RACE analysis has identified two transcription termination signals that generate transcripts of 5.034 and 2.339 kb (24,46), although it is not yet clear whether these transcripts are differentially distributed. In this report we show that alternative promoters regulate CYP4F3 expression in myeloid and hepatic cells. A distinct transcription start site in lymphoid cells suggests that there are at least three tissuespecific promoters in the gene. Regulation of CYP4F3 gene expression is clearly complex and must involve coordinated actions at all levels of mRNA processing to generate the correct isoform in specific locations.