Differentiation-stimulated Activity Binds an ETS-like, Essential Regulatory Element in the Human Promyelocytic defensin-1Promoter*

The human HNP-defensin-1 gene encodes a peptide antibiotic found exclusively in neutrophils and is key to elimination of microbes. Expression is a marker for the granulocytic lineage and for certain stages of differentiation and is not known to be inducible in mature cells under physiological conditions. Low level of transcription also occurs in HL-60 promyelocytic leukemia cells and is greatly activated upon drug-induced granulocytic maturation and by low doses of retinoic acid, in a strictly cell-specific manner (Herwig, S., Su, Q., Ma, Y., and Tempst, P. (1996) Blood 87, 350–364). We have analyzed a 10-kilobase pair region, upstream of thedefensin-1 cap site, for the presence of control elements, and we describe a minimal promoter (position −83 to +82) required to drive transcription in HL-60 cells in a quasi cell-specific manner. Our data also suggest the presence of negative regulatory elements in the −416/−191 region that may further contribute to cell specificity in a chromosomal context. The basal promoter contains two functionally essential, ETS-like (GGAA core sequence) elements. The proximal site (−22/−19) constitutively binds the PU.1 transcription factor in vitro and could function, together perhaps with an adjacent TA-rich sequence (−32/−25), in assembly of a myeloid-restricted, basal transcription factor complex. The distal site (−62/−59) interacts in vitro with an unidentified activity, distinct from PU.1, ETS-1, PEA3, and ELK-1 (factors with definite binding site similarities), and is greatly stimulated by phosphorylation during granulocytic differentiation of HL-60 cells. Identification of this protein will be important to resolve the molecular mechanisms controlling temporal, granulocytic restricted gene expression.

The human HNP-defensin-1 gene encodes a peptide antibiotic found exclusively in neutrophils and is key to elimination of microbes. Expression is a marker for the granulocytic lineage and for certain stages of differentiation and is not known to be inducible in mature cells under physiological conditions. Low level of transcription also occurs in HL-60 promyelocytic leukemia cells and is greatly activated upon drug-induced granulocytic maturation and by low doses of retinoic acid, in a strictly cell-specific manner (Herwig, S., Su, Q., Ma, Y., and Tempst, P. (1996) Blood 87, 350 -364). We have analyzed a 10-kilobase pair region, upstream of the defensin-1 cap site, for the presence of control elements, and we describe a minimal promoter (position ؊83 to ؉82) required to drive transcription in HL-60 cells in a quasi cell-specific manner. Our data also suggest the presence of negative regulatory elements in the ؊416/؊191 region that may further contribute to cell specificity in a chromosomal context. The basal promoter contains two functionally essential, ETS-like (GGAA core sequence) elements. The proximal site (؊22/؊19) constitutively binds the PU.1 transcription factor in vitro and could function, together perhaps with an adjacent TA-rich sequence (؊32/؊25), in assembly of a myeloid-restricted, basal transcription factor complex. The distal site (؊62/ ؊59) interacts in vitro with an unidentified activity, distinct from PU.1, ETS-1, PEA3, and ELK-1 (factors with definite binding site similarities), and is greatly stimulated by phosphorylation during granulocytic differentiation of HL-60 cells. Identification of this protein will be important to resolve the molecular mechanisms controlling temporal, granulocytic restricted gene expression.
Neutrophils are specialized scavenger blood cells, killing microorganisms through a combination of reactive oxidants and polypeptide antibiotics (1). Such peptides are stored in cytoplasmic granules and released, whenever required, into the phagosomes that hold ingested microbes (2)(3)(4). Defensins (also termed "human neutrophil peptides," or HNP) 1 are the major components of this system and account for a large percentage of total granular protein (4 -6). Four different isoforms, HNP-1-4, have been isolated (7)(8)(9), and analysis of cDNA clones has indicated processing from larger precursor structures (10 -13). The mature peptides are 29 -30 amino acids long and are defined by a conserved cysteine backbone (4). "Defensin-like" peptides have also been detected in epithelial linings of the tongue (14), respiratory tract (15), and gut (16,17). Expression of HNP-type defensins is believed to be cell-specific, however, and the non-neutrophilic types are now commonly known under names such as TAP, LAP, cryptdins, and ␤-defensins (18,19).
Even though defensin peptides are abundantly present in differentiated neutrophils, transcripts have never been detected in peripheral blood but rather in unfractionated bone marrow (10,11,20,21). More specifically, transcription seems restricted to a certain window in myeloid blood cell differentiation (11,21). Consistent with these findings is the presence of defensin mRNA, albeit at trace levels, in the HL-60 human promyelocytic leukemia cell line (10,(21)(22)(23)(24)(25). HL-60 cells can be chemically induced to mature along various pathways, thus providing a model system for study of differentiation-specific gene regulation (26 -28). For example, in the course of retinoic acid (RA) treatment, defensin transcription reaches peak levels during the resultant myelocyte and very early metamyelocyte stages of the granulocytic pathway, later followed by a complete down-regulation (25). By contrast, instant down-regulation to virtually undetectable levels was observed during phorbol ester-promoted differentiation toward macrophages (25). Similarly, defensin transcripts have never been found in either myeloblastic (KG-1), monoblastic (U-937), myeloblastic/erythroblastic (K-562), B-lymphoid or T-lymphoid cell lines, not even after extensive RA treatment (10,25). Any studies aimed at understanding this unique granulocytic expression of defensin genes must converge, eventually, at the identification of genomic regulatory elements and their cognate transactivating factors.
Considerable efforts have been expended already at analyzing the control regions of other myeloid-specific genes (29 -31). Instead of being strictly myeloid-specific, many of the transcription factors involved are more commonly expressed, for instance Sp1 (32,33), OCT-1 (33), PU.1 (31,34), PEBP2/CBF (35), myb (36), C/EBP (37,38), and HLH factors (39). Not surprisingly then, lineage-specific gene activation is controlled, in many cases, through unique combinations (30). For example, PU.1 allows Sp1 to bind in a cell-specific fashion (31); likewise, PU.1 together with one or more of C/EBP, AML-1, c-MYB, and HLH factors function as combinatorial activators of myeloid genes (39 -43), as do c-MYB, together with C/EPB or with ETS-1 or -2 (44,45). Furthermore, C/EBP␤ and PU.1 are activated, or have their transactivating potential enhanced, by phosphorylation, which may impart an additional layer of cell specificity (46 -48). Alternative scenarios of myeloid-specific gene activation have ubiquitous factors (e.g. CP1) drive transcription only when promoters are not occupied by repressor proteins (e.g. CDP); here, lineage/stage-specific derepression is the real switch to expression (e.g. in case of gp91-phox) (49). In view of the published data, it is quite possible that defensin transcription in HL-60 cells is also controlled by one or more of the aforementioned transcription factors and repressors. However, inspection of the 1.2-kb upstream sequence and of the first intron of the HNP defensin-1 gene (taken from Ref. 50) did not reveal a presence of the precise binding sites, as previously characterized for these particular factors; neither could RA response elements (51) be identified. Thus, there is no easy way to formulate a mechanistic model for promyelocytic defensin expression at this time. Regulatory elements, and their binding factors, will have to be uncovered and characterized without any preconception of identity.
Here, we describe a minimal defensin-1 promoter, required to drive transcription in HL-60 cells in a quasi cell-specific manner, that contains two essential, ETS-like elements, one binding the PU.1 transcription factor and the other binding a RA-stimulated activity in vitro.

EXPERIMENTAL PROCEDURES
Materials-All-trans-retinoic acid was purchased from Sigma (catalog number R2625); D-luciferin potassium salt was from Analytical Luminescence Laboratory (San Diego, CA), and poly(dI-dC)⅐poly(dI-dC) was from Amersham Pharmacia Biotech. Oligonucleotides were synthesized by the Sloan-Kettering Microchemistry Core facility. Purified TBP (TATA-binding protein) and rabbit polyclonal antibodies, specifically recognizing either PU.1 (sc-352X) or ELK-1 (sc-355X), were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). pCMV-hGH plasmids were a gift from Dr. Daniel Tenen (Harvard Medical School, Boston), and their use in transient transfection assays has been described (34). All other chemicals were from Sigma, unless otherwise indicated.
Cell Lines and Culture Conditions-The human promyelocytic leukemia cell line HL-60, the myeloblastic leukemia cell line KG-1, the monocytic cell line U-937, and erythroleukemic line K-562 were obtained from the American Type Culture Collection (ATCC, Rockville, MD); Burkitt CA-46 lymphoma cells and S3 HeLa carcinoma cells were obtained, respectively, from Drs. A. Zelenetz and J. Hurwitz (Sloan-Kettering, New York); the retinoic acid-resistant cell line HL-60R was provided by Dr. S. Collins (Fred Hutchinson Cancer Center, Seattle, WA). HL-60, HL-60R, U937, K-562, and Burkitt cells were grown in RPMI medium supplemented with 10% heat-inactivated fetal calf serum (FCS) (HyClone Laboratories Inc., Logan, UT), 5.0 units of penicillin, and 5 g/ml streptomycin (complete medium) and maintained at 37°C in a humidified atmosphere containing 5% CO 2 ; HeLa S3 cells were maintained in minimum Eagle's medium Joklin medium (Life Technologies, Inc.) with 5% FCS supplement; KG-1 was cultured in suspension in Iscove's modified Dulbecco's medium (Life Technologies, Inc.) containing 10% FCS and 10 Ϫ4 M ␣-thioglycerol. Cell cultures were always passaged twice a week to maintain a cell density between 2 ϫ 10 5 and 1 ϫ 10 6 cells/ml. Cells were counted in a hemocytometer chamber, and viability was assessed by exclusion of 0.1% trypan blue. For induction experiments, cells were seeded at 2.5 ϫ 10 5 cells/ml; inducers were added 24 h later and were then left in the culture medium for 72 h, unless otherwise indicated. The concentrations of the inducers were as follows: 1 M all-trans-retinoic acid (RA); 160 mM dimethyl sulfoxide (Me 2 SO); 2 mM hexamethylene-bisacetamide (HMBA).
Isolation and Characterization of Genomic Defensin Clones-To allow isolation of genomic clones containing large portions (Ͻ10 kb) of the defensin-1 gene 5Ј-flanking region, polymerase chain reaction (PCR) was first used to generate a defensin-specific probe. Oligonucleotides DEF-38-S (5Ј-AGATACAACCTGACCTGTGTC-3Ј) and DEF-737-AS (5Ј-TCCCGAGGACCTGGGGTCTAACCA-3Ј) were designed based on the published defensin genomic sequence (50) and used as primers. The PCR reaction was done using a Gene Amp System 9600 (Perkin-Elmer), Taq polymerase (Promega, Madison, WI), and 1 g of HL-60 cell genomic DNA as template, 0.2 M oligonucleotide primers, and the following cycling parameters: 94°C for 1 min, 55°C for 1 min, 72°C for 1 min for a total of 30 cycles. The resulting PCR product (722 bp) was then labeled using a random priming labeling kit (Amersham Pharmacia Biotech). Briefly, 50 ng of PCR products in 10 l of nuclease-free water were boiled for 5 min to denature the DNA, followed by submersion of the tube in ice water for 2 min and by adding 10 l of labeling buffer (6 g/m hexadeoxyribonucleotides, 440 mM HEPES, pH 6.6, 110 mM Tris, pH 8.0, 11 mM MgCl 2 , 22 mM ␤-mercaptoethanol, and 44 mM each of dATP, dGTP, and dTTP), 5 l of [␣-32 P]dCTP (NEN Life Science Products), and 5 units of DNA polymerase I Klenow fragment. After 1 h incubation at 37°C, the probe was purified over a Sephadex G-50 column.
This probe was then used to screen a human fibroblast genomic library that had been constructed in the Lambda FIX II vector and was purchased (catalog 946203) from Stratagene (La Jolla, CA). In this way, and under stringent conditions, 21 positives were obtained from approximately 200,000 lambda clones. To confirm that these clones indeed contained defensin genomic sequences, Southern blots were probed with a labeled oligonucleotide, DEF-1216-AS (sequence below), corresponding to exon I (50). Furthermore, to determine which of the clones contained authentic defensin-1 sequences, and not those of the highly similar defensin-3 gene, we screened for the presence of a unique MvaI site within the EcoRI fragment (Ϫ1170 to ϩ425) of defensin-1, a site which is missing from the comparable region in the defensin-3 gene. Of the "MvaI-positive" clones, we selected clone 17 (15-kb insert) for study because it extends the farthest (ϳ10 kb) to the 5Ј-side of the first exon of the defensin-1 gene. This 10-kb fragment was then released by NheI digestion and subcloned into the NheI site of the pGL3 luciferase plasmid vector (Promega).
Primer Extension Assay-Primer extension assays were performed using the appropriate reagent kit from Promega and following the instructions provided by the manufacturer. Antisense primers, DEF-1216-AS (5Ј-CTAGGCAGGGTGACCAGAGA-3Ј) and DEF-2658-AS (5Ј-AGAATGGCAGCAAGGATG-3Ј) (positions indicated on Fig. 2), and X174 HinfI DNA marker (Promega) were separately kinase-labeled with the [␥-32 P]ATP. Fourteen g of total RNA from HL-60 cells was annealed to 0.1 pmol of the labeled primer in the buffer containing 50 mM Tris-HCl (pH 8.3 at 42°C), 50 mM KCl, 10 mM MgCl 2 , 10 mM DTT, 1 mM of each dNTP, and 0.5 mM spermidine. The components were gently mixed, heated to 58°C for 20 min, and then cooled to room temperature for 10 min. The extension reaction was carried out in the buffer described above in the presence of 2.8 mM sodium pyrophosphate and 1 unit of avian myeloblastosis virus reverse transcriptase in a total reaction volume of 20 l. The incubation was performed at 42°C for 30 min. An equal volume of a 2ϫ loading buffer, containing 98% formamide, 10 mM EDTA, 0.1% xylene cyanol, and 0.1% bromphenol blue, was then added, the mixture heated to 90°C for 10 min, cooled on the ice, and a 10-l aliquot loaded onto a 10% polyacrylamide gel (19:1; 1ϫ TBE, containing 8 M urea; 1.0-mm thick), electrophoresed at 250 V constant voltage, at 15°C, until the bromphenol blue marker had reached the bottom of the gel. Dephosphorylated X174 HinfI DNA markers were co-electrophoresed in adjacent lanes as size markers. The gel was then transferred onto Whatman paper, vacuum-dried, and exposed to Hyperfilm (Amersham Pharmacia Biotech) for the desired time at Ϫ80°C with an intensifier screen.
Northern Blot Analysis-RNA extraction, agarose gel electrophoresis, transfer of the RNA to Hybond-N ϩ membranes (Amersham Pharmacia Biotech), and hybridization with a defensin-specific RNA probe were all done exactly as described before (25). Probe template was a 0.45-kb SphI-EcoRI fragment derived from the HNP-1B cDNA clone (10), which was used to generate a [␣-32 P]UTP-labeled RNA, also as described (25). Washed blots were exposed to Hyperfilm-MP (Amersham Pharmacia Biotech), autoradiographs scanned, bands quantitated and normalized to ribosomal RNA levels (determined from ethidium bromide-stained gels), as described (25).
Site-directed Mutagenesis-Mutagenesis of selected nucleotides in the defensin regulatory sequences, contained within plasmid pGL3b-Sau96I (Ϫ83), was done as described by Zaret et al. (52). For the first round of PCR, two pairs of oligonucleotide primers were synthesized for each mutant to be constructed. First, the vector sequence including the restriction site adjacent to the 5Ј-end of the insert of interest was used as the sense primer; the antisense primer was designed from the same region but carrying the nucleotide substitutions. For the second pair, the sense primer was again designed from the same region and with the complementary nucleotide substitution, and the antisense primer was designed from the vector sequence adjacent to the 3Ј-end of the insert including the restriction site. The PCR reactions were performed separately, and their products then used in a second round of PCR by annealing the two overlapping PCR products first, followed by the second reaction which used the sense and antisense primers derived from the vector. The resulting PCR product was then digested with the appropriate restriction enzyme and ligated into the corresponding vector to generate the mutant construct.
Two double mutants were constructed using the Stratagene QuikChange (catalog 200518) site-directed mutagenesis kit as per the manufacturer's instructions. As templates we used the previously single site mutated (two different sites in the proximal promoter region) pGL3b-Sau96I(Ϫ83) plasmids, in conjunction with the appropriate sense and antisense mutation primers (35 nucleotides long) to introduce a GGAA 3 AAGG proximal secondary site modification. All mutagenized plasmid constructs were sequenced to confirm the desired alterations; sequence analysis was done at the DNA Service Laboratory, Biotechnology Center, Utah State University (Logan, UT).
Transient Transfection Assays-Transfection of tissue culture cells and luciferase assays were carried out as described by Pahl et al. (34). In brief, tissue culture cells were diluted into the corresponding growth media (see under "Cell Lines and Culture Conditions"), at densities of 4 ϫ 10 5 /ml, the day before transfection. After 18 h, cells (1 ϫ 10 7 per transfection) were pelleted and washed with pre-warmed (37°C) Iscove's modified Dulbecco's medium, centrifuged at 500 ϫ g for 5 min at room temperature (RT), resuspended at a density of 1 ϫ 10 7 cells in 0.4 ml of warm Iscove's modified Dulbecco's medium containing 2.5 g of pCMV-hGH plasmids. This suspension was added into the electroporation cuvette already containing the luciferase expression DNA constructs (18 g of pGL3-control plasmid in less than 20 l volume; weight amounts of insert-containing plasmids were adjusted to be equimolar with the control). Cells and plasmids were then mixed with a pipette, incubated for 5 min at RT, followed by electroporation at 975 microfarads capacitance and 280 V using a Gene Pulser II (Bio-Rad), unless otherwise indicated. The cells were then transferred to 10 ml of warm Iscove's modified Dulbecco's medium with 10% FCS, the dishes swirled and incubated at 37°C for 5 h, and the cells harvested in 15-ml tubes by centrifugation at 500 ϫ g for 5 min at RT. One ml of supernatant from each experiment was stored in an Eppendorf tube for human growth hormone (hGH) assay (see below). Pellets were washed with 5 ml of phosphate-buffered saline at RT; 300 l of lysis buffer (containing 1% Triton X-100, 25 mM Gly-Gly, pH 7.8, 15 mM MgSO 4 , 4 mM EGTA, pH 7.8, 1 mM DTT) were added, and pellets were then resuspended, transferred to Eppendorf tubes, vortexed for 5 s, and spun at full speed for 3 min at RT. Fifteen l of the above lysate was then mixed with 300 l of freshly made assay buffer, which contained 25 mM Gly-Gly, pH 7.8, 15 mM KPO 4 , pH 7.8, 15 mM MgSO 4 , 4 mM EGTA, pH 7.8, 2 mM ATP, pH 7.8, 1 mM DTT. Relative light units (RLU) were measured for 20 s in a model Monolight 2010 luminometer (Analytical Luminescence Labora-tory, San Diego, CA). The hGH was measured with the enzyme-linked immunosorbent assay kit from the Nichols Institute (San Juan Capistrano, CA) as per the manufacturer's instructions. Briefly, 100 l of supernatant was mixed with an equal volume of antibody solution; the latter is a mixture of two monoclonal antibodies, each one specific for a different and distinct epitope on the hGH molecule, to form a soluble sandwich complex in the presence of hGH. One of the antibodies is 125 I-labeled for detection, whereas the other antibody is coupled to biotin. The reaction was then mixed with an avidin-coated plastic bead and incubated for 90 min at RT while shaking (180 rpm). After two washes, the bead was counted in gamma counter (LKB 1272 Clinigamma; Wallac, Gaithersburg, MD) for 1 min.
Mini-preparation of Nuclear Extract-Nuclear extract was prepared as described (53), with modifications. Briefly, 10 8 cells were pelleted at 500 ϫ g for 5 min at room temperature. The pellet was resuspended in 1.5 ml of ice-cold phosphate-buffered saline and transferred to an Eppendorf tube and spun for 10 s at full speed. The pellet was then resuspended in 1ϫ packed cell volume of cold buffer containing 10 mM HEPES-KOH, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, 0.2 mM phenylmethylsulfonyl fluoride by flicking the tube and leaving it on ice for 15 min. The reaction mixture was then passed five times through a syringe with 23-gauge needle and spun for 20 s at full speed. The supernatant was discarded and the pellet resuspended in two-thirds packed cell volume of ice-cold buffer containing 20 mM HEPES-KOH, pH 7.9, at 4°C, 25% glycerol, 1.5 mM MgCl 2 , 420 mM NaCl, 0.2 mM EDTA, 0.5 mM DTT, 0.2 mM phenylmethylsulfonyl fluoride. After incubating the suspension on ice for 30 min while stirring, it was centrifuged for 5 min at full speed at 4°C. The clear supernatant was then collected, quickly frozen in liquid nitrogen, and stored at Ϫ80°C until further use. Protein concentrations were determined using the Bradford assay (Bio-Rad) and bovine serum albumin standards (Sigma).
Electrophoretic Mobility Shift Assay (EMSA)-EMSA was performed as described by Skalnik et al. (49). In brief, pre-binding of 10 -18 g of the nuclear extract to the poly(dI-dC) was carried out at 30°C for 10 min in buffer containing 4% glycerol, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM DTT, 25 mM NaCl, 10 mM Tris-HCl, pH 7.5, and 0.05 mg/ml poly(dI-dC)⅐poly(dI-dC). In case of competition experiments, the competing oligonucleotides (5-200-fold molar excess) were included in this preincubation mixture. Radiolabeled oligonucleotide probe (3.5 fmol, ϳ2 ϫ 10 4 cpm) was then added to the reaction mixture and incubated at 30°C for 20 min. One microliter of 10ϫ gel loading buffer, containing 250 mM Tris-HCl, pH 7.5, 0.2% bromphenol blue, 0.2% xylene cyanol, and 40% glycerol, was then added to the reaction for loading onto a 4 -6% native gel (which was pre-run for 90 min at 100 V in 0.5ϫ nondenaturing TBE buffer) at 15°C and 125-150 V for about 3 h. The gel was then transferred onto Whatman paper, vacuum-dried, and exposed to Hyperfilm (Amersham Pharmacia Biotech) for the desired time at Ϫ80°C and with an intensifier screen. For antibody "supershift" experiments, antibodies (0.1 g in 1-l volume) were added to the reaction mixtures after the DNA probes had been incubated with nuclear protein for 20 min at 30°C, and the DNA-protein complexes were resolved on 6% polyacrylamide gels, using 0.5ϫ nondenaturing TBE buffer.
Phosphatase Treatment of Nuclear Extract-Potato acid phosphatase (type VII, Sigma) was diluted in 10 mM sodium acetate, pH 5.2, to give concentrations ranging from 0.01 to 0.1 units/l. Eighteen g of nuclear extract (in 1.1-1.4 l volume) was incubated with 1 l of phosphatase for 20 min at 30°C, in the presence of protease inhibitors (20 g/ml each of aprotinin, leupeptin, pepstatin A, and Sigma trypsin inhibitor) and 10 mM Na 3 VO 4 . In negative control experiments, phosphatase was replaced by either phosphate-buffered saline alone or by heat-inactivated phosphatase (10 min at 100°C). Phosphatase-treated nuclear extracts (3 l) were added to 6 l of pre-binding buffer and incubated for 10 min at 30°C; the entire mixture was then mixed with 1 l of labeled probe and incubated 20 min at 30°C (as detailed above under "EMSA").

Defensin-1 Transcriptional Start Site and Promoter Capacity
in HL-60 Cells-Using the published sequence of human neutrophil defensin-1 gene (50), two primers were designed for PCR amplification of a 700-bp fragment, with its 3Ј-end located about 0.5 kb upstream of exon I. The PCR product was used to screen a human fibroblast genomic library; positives were reprobed with a labeled oligonucleotide corresponding to exon I. To determine which of the clones contained authentic defensin-1 sequences, and not those of the highly similar defensin-3 gene, we screened for the presence of a unique MvaI site within the EcoRI fragment (Ϫ1170 to ϩ425), a site which is missing from the corresponding region in the defensin-3 gene. One of the MvaI-positive clones (def-1, number 17) was selected for further study as it contained the entire defensin-1 gene, plus about 10 kb of 5Ј-flanking sequence.
The defensin-1 gene consists of 3 exons, the first one relatively small and coding for 5Ј-untranslated mRNA sequence only; the first intron is 1.4 kb in size (see Fig. 1A; after Linzmeier et al. (50)). The exact location of the major transcriptional start site (defining the 5Ј boundary of exon I) was mapped by primer extension analysis, using HL-60 cell RNA and two antisense oligonucleotides complementary to regions in, respectively, exons I and II (as indicated in Figs. 1A and 2). The extended products resulted in bands of, respectively, 65 and 120 nucleotides long (Fig. 1B), indicating an apparent defensin-1 promoter transcriptional start site as marked in Fig.  2 (designated ϩ1). Mapping of this site was then confirmed by S1 nuclease protection analysis. As expected, when a 145-bp genomic probe (Ϫ63 to ϩ82) was hybridized to HL-60 total RNA, a single 76-bp protected fragment was detected, whereas a control assay with yeast RNA did not show any band (data not shown). It thus appears that exon I is 76 bp in length, extending 38 bp further upstream than previously assumed (50), and that a TAAATA sequence, conserved between several human and rabbit myeloid defensin genes and earlier postulated as a putative TATA box (50), is therefore located at the Ϫ30to Ϫ25-bp position in the current numbering scheme (Fig.  2), consistent with the preferred TATA box region in vertebrates (54). The 5Ј-flanking sequence of the defensin-1 gene was carefully mapped. The sequence was in accordance with the previously reported one, except for three nucleotides as follows: a TA instead of an AT at positions Ϫ551 and Ϫ550, which introduced a ScaI restriction site; and at Ϫ39, where a C was found instead of an A.
By using Northern blot and RNase protection analysis, we and others (10,25) have previously detected the presence of defensin transcripts in HL-60 cells but did not obtain measurable signals in any other myeloid or lymphoid leukemic cell lines. It should be noted, however, that basal levels of defensin mRNA in HL-60 cells are quite low but that a cell-specific, 50 -100-fold induction is achieved upon retinoic acid (RA)-induced differentiation (25). Because RA-dependent defensin activation is a fairly late event (24 -36 h to reach a 10-fold induction), it could not be monitored using transiently transfected reporter constructs, for luciferase activity peaks at 5 h after transfection of uninduced HL-60 cells and then falls off sharply. 2 Alternatively, transfections done 1-2 days after addition of inducer proved to be very inefficient and did not provide a feasible approach for the HL-60 cell system either. Thus, we decided to focus on locating control elements that might govern specific gene expression, however weak, in promyelocytic cells. defensin-1 sequences located upstream of intron I were therefore systematically truncated from Ϫ10 kb to Ϫ552 (and with fixed 3Ј-ends at ϩ82), inserted in the promoterless luciferase reporter vector pGL3-Basic (see Fig. 3A), and tested for their ability to drive transcription after transfection into promyelocytic (HL-60), myeloblastic (KG-1), and monoblastic (U937) leukemia cell lines (nomenclature taken from 2 Y. Ma and P. Tempst, unpublished observations. Ref. 28). As shown in Fig. 3B, the human defensin-1 Ϫ552/ϩ82 promoter (containing 552 bp of 5Ј-flanking sequence, the entire exon I, and 6 bp of the first intron) reproducibly expressed Ͼ30-fold more luciferase activity in HL-60 cells than the promoterless reporter and was about 50% more active than a SV40 promoter in this regard and in these cells. By contrast, a mere The ScaI restriction sites and the corresponding sizes are marked. B, the cell lines (HL-60 promyelocytic, KG-1 myeloblastic, and U-937 monoblastic) and the plasmid constructs, containing gradually 5Ј-truncated sequences (from Ϫ10 kb to Ϫ552; 3Ј-end fixed at ϩ82), used in transient transfection assays are shown on the left; the results of the luciferase expression analyses are shown on the right. Promoter activity, measured as relative light units (RLU), is normalized for transfection efficiency by co-expressing human growth hormone (hGH), driven by a constant CMV promoter and measured by enzyme-linked immunosorbent assay. The results, the mean of at least three experiments, are expressed as RLU per ng hGH and are shown here as percent of luciferase expression driven by an SV40 early promoter (arbitrarily assigned 100% and also normalized by CMV-driven hGH co-expression) in the same cells. 0-Luc indicates promoterless luciferase gene. Further details can be found under "Experimental Procedures" and in the text. 7-fold transcriptional increase over a promoterless reporter and only one-third of the SV40 promoter activity were noted in U937 cells. In KG-1 cells that same defensin-1 promoter sequence (Ϫ552/ϩ82) yielded reproducibly lower luciferase activity than the promoterless reporter alone. It should be noted that in all these, and all following, transient transfection assays, luciferase activity was normalized for transfection efficiency by measuring secreted levels of growth hormone after co-transfection with a plasmid containing the human growth hormone gene under a cytomegalovirus (CMV) promoter control.
The Ϫ83/Ϫ51 Region of the defensin-1 Promoter Contains a Positive Regulatory Element and Binds an RA-stimulated, Nuclear Phosphoprotein from HL-60 Cells-The Ϫ552/ϩ82 defensin-1 genomic DNA fragment then served to generate a series of six 5Ј-truncation products, utilizing either suitable restriction sites or exonuclease III digestion (see "Experimental Procedures"; positions indicated in Fig. 2), which were all inserted in front of the luciferase reporter gene (in pGL3-basic vector) to again assess abilities for activating transcription in vivo. Interestingly, as upstream sequences were systematically deleted from Ϫ552 to Ϫ83, promoter activity in HL-60 cells moderately (by 70%) increased to about 65-fold over the promoterless control (Fig. 4A). Further truncation of the 5Ј-end by another 33 bp (to Ϫ50/ϩ82) resulted in a more than 10-fold reproducible reduction of measured luciferase activity. Clearly, an important positive regulatory element, or elements, must be contained within the Ϫ83/Ϫ51 region of the defensin-1 promoter acting in HL-60 cells. To examine cell specificity of this element, the same constructs were also transiently transfected into different myeloid, myeloblastic/erythroblastic (K-562), and lymphoid (Burkitt B-cell lymphoma) leukemia cell lines and into HeLa carcinoma cells. The results shown in Fig. 4B indicate that, although the capacity of the Ϫ83/ϩ82 defensin-1 promoter to drive transcription in vivo is not entirely exclusive to promyelocytic cells, it certainly is more efficient in HL-60 than in the other cell lines. Whereas its activity exceeds that of an SV40 promoter (arbitrary 100% activity) in HL-60 cells (250%), it is consistently less in U-937/KG-1 (both ϳ80%), K-562 (35%), Burkitt (30%), and HeLa (25%) cells. It appears therefore that Ϫ83/ϩ82 defensin-1 promoter activity is directly correlated to the extent that cell lineage and differentiation stage resemble the promyelocytic phenotype; about 3-fold lower activity was measured in myeloid cells of a lesser (KG-1) or more advanced (U-937) maturation stage, 8-fold lower in Bcells, and about 10-fold lower in non-blood (HeLa) cells. Moreover, defensin transcription is strongly activated during granulocytic differentiation of HL-60 cells but virtually uninducible in any other cell line tested so far (25). Promoter (Ϫ83/ϩ82) leakiness, in terms of cell specificity, may have to do with deletion of upstream negative regulatory elements, as extending the 5Ј-end of the promoter region construct from position Ϫ83 to Ϫ552 resulted in a reduction of in vivo activity by more than 5-fold in KG-1 and Burkitt cells but only by half in HL-60 cells.
To determine whether any nuclear factors might bind to the Ϫ83/Ϫ51 region of the defensin-1 promoter, electrophoretic mobility shift assays (EMSA) were performed using a 30-bp-long (Ϫ83/Ϫ54), double-stranded oligonucleotide (D1 in Fig. 5A) and an overlapping oligonucleotide (D2), spanning the region Ϫ64/ Ϫ35 (just 5Ј to the TA-rich sequence situated at the conventional TATA box position), as probes. We will refer to the contiguous sequence comprised by oligos D1 and D2 as the "D box" (Ϫ83/Ϫ35; shown boxed in Fig. 2). Binding of either probe to nuclear proteins from untreated HL-60 cells was not readily apparent in our initial assays (Fig. 5B); however, longer expo-sures indicated two discrete, faint bands ( Fig. 5C; zero time point). Oligonucleotide probes containing AP-1 and OCT-1 binding sites, on the other hand, yielded weak but clear bands upon short exposure, as previously observed (55,56). When nuclear extracts were prepared from HL-60 cells that had been treated with RA for 72 h, a strong protein complex formed with probe D1, which could be fully competed with self-oligonucleotide, indicating specificity of interaction (Fig. 5B). Likewise, prominent, single band complexes formed between both AP-1 and OCT-1 probes and nuclear proteins of RA-treated cells; compared with the control experiments, these bands were of slower mobility (upward shift) and also different from the D1specific band. By contrast, probe D2 yielded only a very faint band of comparable mobility as the complex involving D1 and a slightly darker band of higher mobility. When a third D boxderived synthetic oligonucleotide (Ϫ74/Ϫ44; overlapping equally with the sequences of probes D1 and D2) was used as an EMSA probe, moderate intensity upper and lower bands were obtained; a pattern that could be considered halfway

FIG. 4. A minimal defensin-1 promoter confers quasi-specific promyelocytic expression in vivo.
A, transient transfection assays for defensin-1 promoter activity in HL-60 promyelocytic cells, and using gradually 5Ј-truncated sequences (from Ϫ552 to Ϫ30; 3Ј-end fixed at ϩ82) inserted in front of the luciferase (Luc) gene. The results are the mean of at least three experiments, carried out as described in Fig. 3 and under "Experimental Procedures." RLU from SV40 promoterdriven expression was normalized per ng of secreted hGH, co-expressed under CMV promoter control, and arbitrarily assigned a value of 100%. 0-Luc indicates a promoterless luciferase gene. B, several of the above constructs were then also transiently transfected in U-937 monoblastic, KG-1 myeloblastic, K-562 erythroleukemic, Burkitt lymphoma (B-cell) and HeLa carcinoma (non-blood cell control) cells; and luciferase expression (per ng of hGH) was monitored as described in A.
between those with probes D1 and D2 (data not shown). We conclude from these results that the major binding site of an RA-inducible factor (or factors) interacting with D box defensin-1 promoter sequences is largely, or fully, contained within region Ϫ83/Ϫ51. Thus, even though we originally determined this region to be essential for in vivo transcription in unstimulated HL-60 cells, it may also contain a regulatory element (perhaps the same) instrumental in RA-enhanced transcriptional activation. Consistent with this concept was the subsequent observation that "induction/stimulation" kinetics of the DNA-binding activity, as measured by EMSA with D1 as a probe (Fig. 5C), closely followed kinetics of transcriptional activation, as measured by nuclear run-on assays, and of steady state defensin mRNA levels as monitored by Northern blotting (25); following addition of RA, a major up-regulation occurred between 24 and 48 h, to continue for at least 1 to 2 more days. Furthermore, the quantitative aspect of D1 binding activity in nuclear extracts of HL-60 cells exposed for 72 h to either RA, HMBA, or Me 2 SO (two compounds previously shown to also induce defensin transcription) was nicely correlated with the effects of these same agents on transcript levels from highest to lowest: RA Ͼ HMBA Ͼ Me 2 SO (Fig. 5C and Ref. 25).
In recognition of the fact that many known transcription factors, including authentic myeloid-active types, become functionally stimulated by phosphorylation (46 -48), we investigated the possibility that granulocytic differentiation-enhanced, D1 binding activity could be the result of such modifications. To this end, nuclear extracts of RA-treated HL-60 cells were subjected to phosphatase treatments of increasing stringency (0.01-0.1 units of potato acid phosphatase type VII per 18 g of protein) prior to EMSA. The results shown in Fig. 5D clearly indicate that the binding activity contained within the upper complex is indeed phosphorylation-dependent, as detection was entirely abolished by active phosphatase but left unchanged when heat-inactivated enzyme was used instead. We do not know the significance of the D1 binding activity in the lower band as its presence was highly variable throughout our studies, compare, for instance, the single band in Fig. 5B (ϩRA lane) with the double bands in Figs. 5, C and D (72 h RA lanes). Although one could speculate that it represents the unphosphorylated form of the upper band, no conclusive experimental evidence is available, since the bottom complex might contain an unphosphorylated, unrelated factor competing for in vitro binding to the D1 oligonucleotide.

A Distal GGAA (Ϫ62/Ϫ59) Sequence in the defensin-1 Promoter Essential for in Vivo Transcription and Interaction with a Putative ETS Family Nuclear Phosphoprotein in HL-60
Cells-To narrow down the sequences in region Ϫ83/Ϫ51 that are (i) essential for transcriptional activation in vivo and (ii) involved in interaction with nuclear proteins in vitro, the effects of selected mutagenesis were analyzed. We also did not know whether the two elements were fully separated, overlapping, or identical. A computer-aided search of the D box sequence for the presence of transcription factor binding sites, using the MatInspector algorithm (57), indicated a core consensus sequence (GGA(A/T)) found within the binding sites of several members of the ETS family of transcription factors, such as ETS-1, ELK-1, and PU.1 (58 -60). Thus, we synthesized mutant D1 oligonucleotide probes, having 5Ј-GGAA-3Ј replaced by AAGG (labeled D1M2 in Fig. 6A), and we constructed luciferase reporter plasmids containing a similarly modified Ϫ83/ϩ82 defensin-1 promoter region (Fig. 6C). Likewise, two additional mutant oligonucleotide probes, and associated mutant reporter constructs, were generated. GGAA

FIG. 5. Differentiation-stimulated, nuclear phosphoprotein from HL-60 cells binds D1 probe (؊83/؊54 of defensin-1 gene) in vitro.
A, schematic representation of two overlapping, double-stranded oligonucleotide (D1 and D2) probes (solid bars) and their relative positions within the D box (Ϫ83 to Ϫ35) of the defensin-1 promoter sequence. The sequences are listed under "Experimental Procedures." B, D1 and D2 were utilized as labeled probes in EMSA experiments, using nuclear extracts from HL-60 cells that had either not (Ϫ) been treated or treated (ϩ) for 72 h with 1 M retinoic acid (RA). Self-competition with a 200ϫ excess of unlabeled oligonucleotide is indicated in the "competitor" row by ϩ. Oligonucleotide probes containing AP1 and OCT-1 transcription factor binding sites were used as controls. Full experimental details are to be found under "Experimental Procedures." C, EMSA with the D1 probe as described above, using nuclear extracts from HL-60 cells treated for various periods (0 -72 h) with 1 M RA; or for 72 h with either HMBA or Me 2 SO. D, EMSA with the D1 probe, using nuclear extracts from 1 M RA-treated (72 h) HL-60 cells. Extracts were either used directly or after further treatment with 0.08 units (U) phosphatase (PPtase; potato acid phosphatase type VII) in the presence of protease inhibitors and phosphotyrosine phosphatase inhibitor. Heat-denatured phosphatase (PPtase ϩ heat) was used as a negative control for enzymatic activity.
(Ϫ62/Ϫ59) replacement in the D1 probe completely abolished formation of the upper complex in gel shifts using nuclear extracts of RA-treated HL-60 cells (Fig. 6B). The exact same mutation resulted in a 12-fold reduction of in vivo transcriptional activity after transient transfection in untreated cells, the equivalent effect of deleting the entire D box (Ϫ83/Ϫ35) region (Fig. 6C). Replacement of a slightly more 5Ј-located tetranucleotide (Ϫ70/Ϫ67), on the other hand, resulted in the loss of the lower band from the EMSA pattern, and in a less than 3-fold reduction of luciferase activity after transfection (Fig. 6, B and C). D box mutations just downstream from the GGAA sequence in the D2 probe and in reporter constructs, namely of tetranucleotides (Ϫ54/Ϫ51), did not result in attenuation of nuclear factor binding nor of in vivo transcriptional capacity. In fact, the single, lower band in EMSA was more intense, and luciferase activity was also slightly increased as compared with the D2 probe and wild type promoter sequence controls, respectively (Fig. 6, B and C). Transient transfection of all the above mentioned constructs in untreated KG-1 cells resulted in similar trends of decreased, and increased, normalized luciferase activity. As can also be seen in Fig. 6C, constructs (Ϫ34/ϩ82) missing the entire D box, or containing a promoterless luciferase gene, yielded reproducibly higher luciferase activities when transfected in KG-1 than in HL-60 cells, for reasons unknown to us at this time. In sum then, the GGAA (Ϫ62/Ϫ59) tetranucleotide sequence in the defensin-1 promoter is essential for in vivo transcription and in vitro binding of RA-inducible nuclear factor(s) in HL-60 cells. Because another GGAA sequence is located between the D box and the first exon of the defensin-1 gene, at position Ϫ22/Ϫ19, we will refer to the 5Ј-most located one as the "distal GGAA." We then sought to determine the identity of the nuclear factor(s) binding to the distal GGAA sequence. Because the transcription factor PU.1 is expressed in HL-60 cells and has been implicated in the regulation of many myeloid-specific genes, it was a likely candidate (31,61). Thus, we tested whether an oligonucleotide ("PU1"), containing a characterized PU.1 binding site from the SV40 enhancer (60,62), could compete the protein(s) binding to the defensin-1 promoter-derived D1 probe. As shown in Fig. 7A, PU1 did not compete binding to D1 probe as efficiently as D1 did in the self-competition experiment. In the reversed experiment, a 200-fold excess of D1 did not at all compete binding to a PU1 probe in gel shift assays using the same RA-treated HL-60 cell nuclear extracts; PU1 self-competition was quite evident and also confirmed that the complexes involving PU.1 transcription factor migrated faster (lower position) on the gel. Complexes containing PU1 probe could also be supershifted using an anti-PU.1 antibody, whereas those containing D1 probe could not (Fig. 7B). The data all but eliminate the possibility that PU.1 binds the distal GGAA (Ϫ62/Ϫ59) in the defensin-1 promoter neither by itself nor as part of a possible multi-protein complex binding the GGAA-containing D1 oligonucleotide in vitro.
Examination of the "ETS protein" literature indicated two other previously described transcription factor binding sites with substantial homology to the CCCCGGAAATT (Ϫ66/Ϫ56) region in the D1 probe sequence, namely the ETS-1/PEA3 consensus binding site ("ETS") and an ELK-1 autonomous DNA-binding site ("ELK") (59,63). Hence, we tested oligonucleotides containing those sequences for their ability to prevent  Fig. 3 and under "Experimental Procedures." RLU from SV40 promoter-driven expression was normalized per ng of secreted hGH, co-expressed under CMV-promoter control, and arbitrarily assigned a value of 100%. Luc. indicates a promoterless luciferase gene.
formation of complexes between a D1 probe and nuclear proteins from RA-treated HL-60 cells. From the results of EMSA experiments shown in Fig. 7C, it appeared that competition for binding to the D1 probe by a 200-fold excess of ETS oligonucleotide was incomplete, albeit better than PU1 competition. In keeping with these observations, gel shifts using a labeled ETS probe (21 bp) resulted in multiple complexes of clearly different mobility, which could not be explained by the different probe sizes alone (Fig. 7C). Complex formation of the ELK probe itself and competition with the D1 probe for binding raised the possibility that ELK-1 transcription factor might, in fact, bind to the Ϫ66/Ϫ56 region of the defensin-1 promoter in HL-60 cells (data not shown). It should be noted, however, that the sequence contained within the ELK probe corresponds to an autonomous binding site for purified ELK-1 (59). Autonomous DNA binding (i.e. external from a ternary complex with the serum response factor) has not yet been shown to occur in whole nuclear extract. Moreover, no serum response element with a consensus CArG box (64) can be located within reasonable distance from the defensin-1 promoter distal GGAA. Thus, the factor(s) contained within the D1 complex could be a novel member(s) of the ETS family of transcription factors.
To investigate cell-specific aspects of this binding activity, the D1 probe was then incubated with proteins from various other cell types and complex formation monitored by gel shift analysis. Interestingly, nuclear proteins of untreated KG-1 and K-562 cells specifically interacted with the D1 probe as well, to yield bands of equal mobility as those in RA-treated HL-60 cells (Fig. 8, A and B); these activities were not further enhanced by RA treatment. Moderate D1 binding activity was stimulated by RA in U-937 cells, but the resulting upper band was consistently of much lower intensity than its counterpart in induced HL-60 cells (Figs. 5, B-D, and 6B). Subsequent analysis indicated that, in contrast to the RA-stimulated/HL-60 derived factor(s), constitutive D1 binding activity from those other cells could not be entirely abolished by similar phosphatase treatments (Figs. 8B versus 5D), suggestive of possibly distinct activities with shared DNA-binding properties in vitro. This hypothesis was further investigated by comparative EMSAs (RA-treated HL-60 versus untreated KG-1 cells) using 10 different single point mutation (at positions Ϫ65 to Ϫ56 in the defensin-1 promoter sequence) derivatives of the D1 oligonucleotide probe (Fig. 8C). We selected that particular region of D1 for mutagenesis because the GGAA core consensus se- quence for recognition by ETS family proteins is at its center location. Not surprisingly then, replacing either one of those four core nucleotides (and also of the 5Ј-adjacent C) completely abolished incorporation of the D1 oligonucleotide in a HL-60 protein-containing complex (Fig. 8C); exchange of either the upstream C (Ϫ64) at the 5Ј-edge of this CGGAA essential sequence or of the A (Ϫ58) or G (Ϫ57) just downstream from it resulted in diminished binding as judged from less incorporation of labeled mutant probe in the band. By contrast, interaction of the same mutant probes with KG-1 nuclear proteins showed some marked differences in EMSA banding patterns, most notably in the ability of D1 mutant "GGAA 3 GGAC" to still incorporate in a prominent, slow-migrating complex (Fig.  8). Overall, the band shift patterns point at subtle differences in preferred DNA-binding sites for HL-60 and KG-1 D1-binding proteins, again implicating likely differences in identity.
Proximal GGAA (Ϫ22/Ϫ19) and TA-rich (Ϫ32/Ϫ25) Sequences in the defensin-1 Promoter Implicated in Transcriptional Activity in HL-60 Cells and in Vitro Binding of PU.1-Even though the ability of the defensin-1 promoter (Ϫ83/ϩ82) to activate transcription in transiently transfected HL-60 cells was severely impaired when D box sequences (Ϫ83/Ϫ35) were deleted, we still measured luciferase activities on the order of 25% of SV40 promoter-driven transcription (Fig. 6C). However, the remaining activity was almost entirely lost upon further deletion of promoter 5Ј-sequences to position ϩ11 (Fig. 9B). Inspection of the Ϫ35/ϩ82 region indicated a TA-rich sequence TTTAAATA (Ϫ32/Ϫ25), already postulated as a candidate TATA box, and of a second GGAA sequence (Ϫ22/Ϫ19), from here on referred to as the "proximal GGAA." To demonstrate possible functional importance of the TA box and the proximal GGAA, two separate trinucleotide mutations were introduced in the Ϫ83/ϩ82 promoter, and the changes of in vivo transcriptional activity was assessed. The combination of three point mutations (A Ϫ29 3 C, A Ϫ28 3 T, and T Ϫ26 3 G) caused a 5-fold reduction in promoter activity, and changing of GAA (Ϫ21/Ϫ19) to CCC resulted in a 2.5-fold decrease. Significantly, when assessed in combination with a mutated distal "GGAA" site (see Figs. 6C and 9B), the effect of the proximal AAAT 3 CTAG modification was a total loss (Ͼ50-fold reduction; equal to negative control) of in vivo promoter function in HL-60 cells (Fig.  9B).
Similar proximal trinucleotide substitutions were then also introduced in a 30-bp oligonucleotide (Ϫ39/Ϫ10), termed "TA" for the inclusion of the TA box (Fig. 10A), and comparatively tested by EMSA for potential differences in nuclear protein binding. After incubation with nuclear extract from RA-treated HL-60 cells, wild type TA probe shifted two distinct complexes, one migrating a little higher than the D1 complex and a second multi-band complex, located lower on the gel, that appeared very similar to a PU1 complex ( Fig. 10B; and Fig. 7, under PU1). Both complexes were easily competed with self-TA oligonucleotide, whereas PU1 and ETS oligonucleotides (the same as used in the experiments shown in Fig. 7) specifically competed out just the lower complex (Fig. 10B); D1 and ELK oligonucleotides were less effective in this regard. Addition of specific anti-PU.1 antibodies also resulted in the complete disappearance of the lower complex further supporting the notion that PU.1 transcription factor, from HL-60 nuclei, does indeed bind to the TA probe.
Interestingly, no such PU.1-TA probe complex was obtained with nuclear extracts from KG-1 cells, indicating that active PU.1 may be absent from these cells (Fig. 10C). This was confirmed by probing KG-1 nuclear proteins with a specific PU.1-binding oligonucleotide (same as used in gel shifts shown in Fig. 7); in the same way, PU.1 was also found absent from HeLa, K-562, and Burkitt cells and was barely detectable in U-937 cells (Fig. 10C). Comparison of the TA probe to the PU1 probe EMSA patterns also indicated that more HL-60 nuclear proteins than just the PU.1 transcription factor bind to the Ϫ39/Ϫ10 region in the defensin-1 promoter in vitro. As for the possible identity of this (these) additional protein(s) involved in upper EMSA complex, judging from the competition experiments it seems unlikely that any of the ETS family members tested here would be involved, including the putative novel factor binding to the D1 probe. Single point mutational analysis of the TA probe (individual base pair changes from positions Ϫ31 to Ϫ18) also indicated that the nucleotides critical for binding of PU.1 and of the unidentified factor, while overlapping, are unquestionably different. Predictably, mutant oligonucleotides TAM1 (modification in TA box) and TAM2 (modification in GGAA), when used as probes, both failed to form PU.1 complexes, and the TAM2 probe did not form an upper complex either (Fig. 10B). Accordingly, 200-fold excess of mutant TAM2 oligonucleotide did not prevent any HL-60 nuclear proteins from binding the wild type TA probe. Excess TAM1 oligonucleotide, on the other hand, fully competed out TA probe, including formation of the PU.1 complex, even though a TAM1 probe did not shift PU.1 by itself. Since repeated experiments confirmed these findings, we assume that the discrepancy may be related to the large excess of mutant oligonucleotide used for competition. In an attempt to address the question whether the defensin-1 promoter TTTAAATA sequence (Ϫ32/ Ϫ25), already shown to be functionally important in vivo, could bind the general transcription factor TFIID complex, we used an excess oligonucleotide containing a bona fide TATA box sequence (termed "TFIID") as competitor. Although the TFIID oligonucleotide, when used as an EMSA probe, bound purified TATA-binding protein (TBP) to some extent (data not shown), it failed to compete for binding of HL-60 proteins to the TA probe (Fig. 10B). Neither did purified TBP bind to the TA oligonucleotide in vitro (data not shown).

DISCUSSION
Cell differentiation involves the orderly expression of numerous genes. Among the transcription factors involved are those governing differentiation itself and others regulating promoters of genes specific for the functions of mature cells. Defensin antibiotics are such key components in neutrophils. In fact, human HNP-type defensins are found exclusively in these cells. Contrary to the insect immune system (65), HNP defensin synthesis is not noticeably induced by microbial challenge but, instead, occurs in maturing bone marrow cells. Its expression is therefore uniquely restricted; it is a genuine marker for neutrophilic lineage and for differentiation stage. As acute myelogenous leukemia cells derive from a block in differentiation, they provide a freeze-frame to analyze molecular events at a given stage. Low level defensin transcription occurs in HL-60 promyelocytic cells but is greatly activated upon (i) drug-induced granulocytic differentiation and (ii) treatment with retinoic acid at doses too low to bring about morphological changes (25). The underlying molecular mechanisms of the basal expression and of the activation processes are unknown. Thus, we sought to analyze the molecular control of defensin transcription and how it is affected by chemotherapeutic drugs. This information may also provide novel insights into the pathogenesis of promyelocytic leukemia and may lead to a better mechanistic understanding of chemotherapeutic interventions as well.
Here, we show the capacity of extended defensin-1 promoter regions, located at Ϫ552/ϩ82, to drive gene expression in vivo in a quasi cell-specific manner; promoter activity in promyelocytic (HL-60) cells is 5-30-fold higher than in related monoblastic (U-937) and, respectively, myeloblastic (KG-1) cell lines. Truncation from position Ϫ552 onward to a minimal promoter, located at Ϫ83/ϩ82, results in enhanced transcriptional activity in HL-60 cells and even more so in other cells, suggesting deletion of negative regulatory elements that may contribute to cell-specific expression in a regular chromosomal context. The existence of possible repressors binding to those elements was not studied further here. Instead, we provide data that demonstrate the presence (and near location) of two positive regulatory elements within the 83-base pair minimal region required for defensin-1 expression in HL-60 cells. Functional significance is implicit from the fact that disruption of the proximal or distal sites results in, respectively, a 5-or 10-fold loss of basal promoter activity; more than that, the double knock-out entirely abolishes all function.
The proximal control element contains a TA-rich sequence TTTAAATA (Ϫ32/Ϫ25) that, by established criteria of weighted consensus sequence and location (54), fulfills all requirements of a vertebrate "TATA box." A trinucleotide mutation (TT-TCTAGA), created within this limited sequence, had a profound negative effect on minimal (Ϫ83/ϩ82) promoter activity in promyelocytic cells, consistent with the predicted functional role. However, an oligonucleotide containing a bona fide TATA FIG. 9. Proximal TA-rich (؊32/؊25) and GGAA (؊22/؊19) sequences in the defensin-1 promoter are functionally important for promyelocytic expression in vivo. Transient transfection assays for defensin-1 minimal (Ϫ83/ϩ82) promoter activity in HL-60 promyelocytic cells. The wild type and variously mutated (singly or doubly; shown below the wild type sequence in A) promoter sequences were inserted in front of the luciferase (Luc) gene. The transcriptional start site is indicated by an arrow, and the box represents the first exon, as labeled (A). The left side portion of B represents the plasmid constructs used in the transfection assays; the solid bar represents the wild type sequence, and the empty areas represent the triple or quadruple base substitutions. The boxes indicate the first exon of defensin-1, and those open on the right represent the reporter gene, labeled Luc. The constructs labeled Ϫ34 and ϩ11 contain defensin-1 promoter sequences located at positions Ϫ34/ϩ82 and ϩ11/ϩ82 (missing 10 bp of exon 1), respectively. The results are the mean of at least three experiments, carried out as described in Fig. 3 and under "Experimental Procedures." RLU from SV40 promoter-driven expression was normalized per ng of secreted hGH, co-expressed under CMV promoter control, and arbitrarily assigned a value of 100%. box sequence failed to compete for binding of HL-60 nuclear proteins to an EMSA probe comprising this TA-rich regulatory sequence ("TA probe"; Ϫ39/Ϫ10), raising doubts about the precise role in defensin-1 expression. Two sets of complexes formed with proteins from HL-60 and U-937 cells. The faster migrating "ladder" complex almost certainly contains transcription factor PU.1, binding to the GGAA core sequence (Ϫ22/Ϫ19) just downstream from the TA site; its banding pattern in mature myeloid cells is the result of differential phosphorylation or represents degradation products (69). Since PU.1 expression is limited to myeloid and B-cells (31,33,61,66,67), it explains the absence of this complex from HeLa cells; early myeloid KG-1 cells could be either too immature to express PU.1 or may have other disregulated expression or activity, perhaps a reason for failure to differentiate (68). The PU.1-binding site is functionally active as well, since its disruption resulted in reduced promoter efficacy in vivo. It appears therefore that defensin-1 is yet another myeloid gene target regulated, at least in part, by PU.1 (31). This brings into perspective the possibility that defensin-1 could indeed have a TATA-less myeloid promoter, relying instead on the proven capacity of nearby bound PU.1, and possible associated proteins, to tether TBP and TFIID and assemble a basal transcription factor complex within reasonable distance from the cap site (70,71). If indeed the case, it still cannot alone account for granulocytic specificity of defensin-1 expression.
Additional control is most likely exerted by a more distal element, also containing the ETS family GGAA signature binding site (Ϫ62/Ϫ59; see Fig. 2). A mere switch of the sequence within this tetranucleotide (to AAGG) caused a substantial loss of minimal (Ϫ83/ϩ82) promoter activity in vivo, the same reduction, in fact, that resulted from truncating the 5Ј-end by 53 bp (to Ϫ30/ϩ82) and arguing for major functional significance. We have termed this distal positive regulatory site REDE (regulatory element of defensin expression). A minor EMSA complex formed with an REDE-containing oligonucleotide (D1 or REDE probe) but markedly gained in intensity after HL-60 cells had been treated with RA or HMBA for at least 2 days, indicating differentiation stimulated binding activity, largely the result of phosphorylation. The GG 7 AA nucleotide exchange, when introduced into this probe, abrogated all binding of HL-60 nuclear protein in vitro. A second, faster migrating band was sometimes observed and may represent the unphosphorylated (or underphosphorylated) form of the major REDE binding activity. The phosphorylation-dependent interaction with specific nucleic acid sequences has broader implications for possible factor identity since several ETS family members require RAS-dependent phosphorylation to reach optimal transactivation potential (72,73), whereas others are negatively regulated in this manner (74,75). Our data convincingly rule out that either PU.1 or ETS-1 interact with the wild type REDE probe, and we speculate, on theoretical grounds, that ELK-1 is absent from the complex as well. Indeed, no serum response element can be located within reasonable distance from REDE, suggesting that if ELK-1 binds to REDE it must do so autonomously and not in its typical capacity of ternary complex factor (59,64,76).
This leaves two key questions: (i) what is the precise identity of this activity, and (ii) how does it get stimulated during granulocytic differentiation? Although the first question can be approached experimentally by analysis of all known ETS family members for REDE binding and for abilities to drive transcription of reporter constructs in non-myeloid cells (after coectopic expression), a more direct and unbiased way is by purification and structural chemical analysis. Until then, we will refer to this activity as IRD (increased REDE binding during differentiation). We know for sure that IRD is different from the protein incorporated in the low mobility EMSA complex that forms with the proximal TA probe, for that band is obtained with extracts from untreated HL-60 cells and cannot be competed with an excess of cold REDE probe. Nuclear proteins from one lymphoid and a few myeloid cell lines formed identical complexes with the same REDE probe in the absence of drug treatment; exposure to RA did not significantly alter those patterns. For lack of IRD-specific reagents, we cannot yet determine whether the same activity is constitutively present in all those cells. However, several lines of evidence indicate that this may actually not be the case. Contrary to the heretofore implicated HL-60 nuclear phosphoproteins, binding activities from the other cells could not be readily eliminated by phosphatase treatments. Furthermore, various band shift patterns resulting from point mutational analyses of the REDE probe indicated subtle differences in preferred nucleic acidbinding sites between proteins from myelocytic and other cells. 1 Ab) was present (ϩ) in the binding reaction of nuclear protein to the TA probe. Oligonucleotides termed D1, PU1, ELK, and ETS have been described in Figs. 5, 7, and 8 and in the text; the sequence of oligonucleotide TFIID, containing a bona fide TATA box, is given under "Experimental Procedures." C, EMSA with labeled TA (see A) or PU.1 (see Fig. 7) probes, using nuclear extracts from untreated (no RA) KG-1, K-562, U-937, and Burkitt lymphoma cells. D, EMSA with a defensin-1 TA oligonucleotide probe and single point mutation (at positions Ϫ32 to Ϫ18) derivatives thereof (total of 15 mutant probes), incubated with nuclear extract from untreated HL-60 promyelocytic or KG-1 myeloblastic cells. The specific nucleotide exchanges, each one characteristic for a particular mutant probe, are indicated (wild type 3 changed to).
Assuming that IRD has transactivating ability in HL-60 cells, it could be that those REDE-binding proteins from other cell types are inactive forms of IRD, for reasons of differential modifications and/or conformational changes (not affecting DNA binding). Alternatively, repressors recognizing the REDE sequence may bind to the probe in vitro and perhaps antagonize putative activity of IRD in vivo by occupying the recognition site in the defensin promoter. In previous reports, ETS repressor factors have already been shown to specifically suppress ETS transcription factor associated activities (77). Yet another possibility is the in vivo incorporation of a suppressor into a bigger complex, also containing bona fide IRD, that might prove unstable during gel shift analysis. And finally, a functional IRD-REDE combination may still fail to drive transcription without a requisite co-regulator, most likely PU.1, bound to the proximal regulatory site, an argument applicable to KG-1 cells for instance.
Regarding defensin and IRD induction by RA or HMBA in promyelocytes, we established a direct correlation over a 3-day period between (i) transcriptional rates and steady state mRNA levels in vivo (25) and (ii) density of IRD binding to the REDE probe in vitro (this study), suggesting a scenario in which the appearance of nuclear IRD might be the final stage of drug-induced signaling toward defensin-1 transcriptional activation, most likely involving activation of a protein kinase. This hypothesis is consistent with an earlier view renouncing granulocytic expression of this gene as the simple consequence of a direct, ligand-mediated RA receptor (RAR) binding to specific elements in its control region but instead predicting a role for RAR more upstream in the pathway (25). Retinoids must also partake to some extent in basal transcription, for the minimal promoter functions 5-fold less effectively 2 in an HL-60 cell subclone, HL-60R, that harbors a dominant-negative RAR␣ mutation and is completely resistant to RA-and HMBA-dependent induction of both differentiation and defensin transcription (25,78). As expected, basal IRD activity was not RA-inducible in these mutant cells either (data not shown).
In summary, defensin-1 expression in promyelocytic leukemia cells is very likely controlled by a combination of at least two nuclear proteins, acting through two functional, ETS-like elements in the minimal promoter. The proximal element binds PU.1 in vitro, and the distal one contains a binding site for a differentiation-stimulated, possibly novel phosphoprotein IRD. Identification and molecular characterization of IRD will be crucial to resolve the molecular mechanisms of defensin-1 activation during drug-induced differentiation and perhaps also during normal granulopoiesis. Protein structural chemistry and mass spectrometric analysis will likely feature prominent roles in achieving the first of these goals.