Macrophage-restricted and interferon gamma-inducible expression of the allograft inflammatory factor-1 gene requires Pu.1.

Expression of allograft inflammatory factor-1 (Aif-1), a 17-kDa protein bearing an EF-hand Ca(2+) binding motif, increases markedly in monocytes and macrophages participating in allo- and autoimmune reactions, including the perivascular inflammation in transplanted hearts, microglial infiltrates in experimental autoimmune neuritis, and the inflamed pancreas of prediabetic BB rats. To investigate the mechanism of this regulation, we isolated the mouse aif-1 gene and determined its genomic organization. The gene has six exons distributed over 1.6 kilobases, an interferon gamma-inducible DNase I-hypersensitive site near -900, and flanking sequences on either side predicted to associate with nuclear matrix. Reporter gene analyses identified sequences between -902 and -789, including consensus Ets and interferon regulatory factor elements, required for macrophage-specific and interferon gamma-inducible transcriptional activity. Pu.1 bound to the Ets site in electromobility shift assay and forced expression of Pu.1 activated the aif-1 promoter in 3T3 fibroblasts, in which it is normally inactive. However, the transcriptional activity of a concatamer of the Ets site alone did not increase with interferon gamma treatment. Cooperation between Pu.1 and proteins binding to the interferon regulatory factor element appears to be necessary for both macrophage-specific and interferon gamma-inducible expression of the aif-1 gene.

Expression of allograft inflammatory factor-1 (Aif-1), a 17-kDa protein bearing an EF-hand Ca 2؉ binding motif, increases markedly in monocytes and macrophages participating in allo-and autoimmune reactions, including the perivascular inflammation in transplanted hearts, microglial infiltrates in experimental autoimmune neuritis, and the inflamed pancreas of prediabetic BB rats. To investigate the mechanism of this regulation, we isolated the mouse aif-1 gene and determined its genomic organization. The gene has six exons distributed over 1.6 kilobases, an interferon ␥-inducible DNase I-hypersensitive site near ؊900, and flanking sequences on either side predicted to associate with nuclear matrix. Reporter gene analyses identified sequences between ؊902 and ؊789, including consensus Ets and interferon regulatory factor elements, required for macrophagespecific and interferon ␥-inducible transcriptional activity. Pu.1 bound to the Ets site in electromobility shift assay and forced expression of Pu.1 activated the aif-1 promoter in 3T3 fibroblasts, in which it is normally inactive. However, the transcriptional activity of a concatamer of the Ets site alone did not increase with interferon ␥ treatment. Cooperation between Pu.1 and proteins binding to the interferon regulatory factor element appears to be necessary for both macrophagespecific and interferon ␥-inducible expression of the aif-1 gene.
Allograft inflammatory factor-1 (Aif-1) 1 is a 17-kDa protein first described in differential mRNA display analysis of chronically rejecting transplanted rat hearts (1,2). The Aif-1 protein was identified in infiltrating mononuclear cells in the allografted hearts, and Western blotting indicated its expression in bone marrow-derived macrophages but not T cells. Gene products with a high level of similarity to Aif-1 have been identified in several other model systems and have been re-ported with the alternative names Iba1 (3), daintain (5), and Mrf-1 (4). Despite the different screening approaches and systems employed in each case, the cells expressing the aif-1 gene product have been derivatives of the monocyte/macrophage lineage; iba1 was reported from differential mRNA display analysis of brain cell cultures, with induction in microglial cells cultured for extended periods of time (3); daintain was identified because of its effects on insulin release and found to localize to microglia in the central nervous system, and macrophages and dendritic cells in other organs (5); and mrf-1, identified in a differential hybridization screen of cerebellar cultures, localized to OX-42-positive microglia in brain sections, with increased expression in activated microglia surrounding injured central motor neurons (4).
The cDNA sequences described in these reports are quite similar but show some variability at the 5Ј end. aif-1 and iba1 sequences diverge at the 5Ј end, consistent with alternative transcription start sites from the same gene (4), but are practically identical downstream of aif-1 base 96 (iba1 base 279). The sequence of daintain, determined from purified porcine protein, is 91% similar to that predicted for rat aif-1 despite the difference in species (5). The 5Ј end of rat mrf-1 is 26 bases longer than that previously reported for aif-1 (1), but the sequence is otherwise the same (4), consistent with the variation in the site of transcription initiation or incomplete 5Ј extension of the cDNA reverse transcription products.
In addition to its expression in mononuclear cells infiltrating cardiac allografts (2,6) and microglia surrounding injured motor neurons (4), Aif-1 protein is increased in microglial cells in models of autoimmune encephalitis (7) and in macrophages infiltrating the inflamed pancreas in a rat model of diabetes (5). Although its cellular function is currently unknown, Aif-1 may be important for monocyte/macrophage effector functions that contribute to inflammation. To begin to understand the molecular basis of the macrophage-restricted and disease-associated pattern of aif-1 expression, we have determined the genomic organization of the aif-1 gene. In addition, we have performed an initial characterization of its promoter and defined regulatory elements by DNase I hypersensitivity, deletion analysis, and site-directed mutation that appear to be essential for its characteristic macrophage-restricted expression. We find that the Ets family transcription factor, Pu.1, and proteins interacting with a site meeting criteria for interferon regulatory factor (IRF) binding are required for expression of the aif-1 gene in both basal and interferon-␥ (IFN-␥)-stimulated states.

EXPERIMENTAL PROCEDURES
Cell Culture and Reagents-RAW264.7 cells and 3T3 fibroblasts (American Type Culture Collection) were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillin (100 units/ml), and streptomycin (100 mg/ml). Cells were seeded at a density of 0.2-0.3 ϫ 10 6 cells/ml. Cytokines were stored at Ϫ80°C and reconstituted or diluted as recommended by the suppliers. Recombinant murine IFN-␥ was purchased from Invitrogen. Recombinant human platelet-derived growth factor BB, IL-1␤, IL-4, and IL-10 were obtained from Collaborative Biomedical, recombinant macrophage colony stimulating factor was from Calbiochem, and recombinant human tumor necrosis factor-␣ was from Genzyme. 12-O-Tetradecanoylphorbol-13acetate was purchased from Sigma. Concentrations are indicated in the figure legends.
Isolation of Genomic Clones-A 489-bp aif-1 cDNA fragment was generated from murine total RNA derived from RAW264.7 cells by the reverse transcriptase PCR (8). Primer sequences, derived from the sequence found under GenBank TM accession number AF074959 were 5Ј-CCAGCCTAAGACAACCAGC-3Ј (forward primer) and 5Ј-ACATC-CACCTCCAATCAG-3Ј (reverse primer). This fragment was radiolabeled with [ 32 P]dCTP and used to screen a mouse (strain 129SvJ) genomic DNA phage library in the vector FixII (Stratagene) as described (9). Hybridizing clones were isolated and purified, and phage DNA was prepared according to standard procedures (9).
RNA Blot Hybridization-Total RNA was obtained from cultured cells by guanidinium isothiocyanate extraction followed by centrifugation through cesium chloride (8). The RNA was fractionated on 1.2% formaldehyde-agarose gels and transferred to nitrocellulose filters, which were then hybridized at 68°C for 2 h to the random-primed, 32 P-labeled mouse aif-1 cDNA probe (10 6 cpm/ml) in QuikHyb solution (Stratagene). The filters were washed in 30 mM sodium chloride, 3 mM sodium citrate, and 0.1% SDS at 55°C and autoradiographed with Kodak XAR film for 48 h at Ϫ80°C. A 32 P-end-labeled oligonucleotide (10) complementary to 28 S ribosomal RNA was hybridized to the filters to correct for differences in RNA loading. Phosphor screens were scanned, and radioactive signal intensity was determined with the ImageQuant v1.1 software analysis program (Molecular Dynamics).
Primer Extension Analysis-Primer extension analysis was performed as described (8). A synthetic oligonucleotide primer (5Ј-ATC-CCTGCTTTGGCTCAT-3Ј) complementary to the 5Ј end of the mouse aif-1 cDNA was end-labeled with [␥-32 P]ATP and hybridized to 10 g of each RNA sample; annealed samples were subjected to reverse transcription. Extension products were analyzed by electrophoresis on an 8% denaturing polyacrylamide gel. The same primer, unlabeled, was used in a sequencing reaction with 33 P-labeled dideoxynucleotide termination (Thermosequenase, Amersham Biosciences) to display the genomic sequence corresponding to the end of the primer extension product. Rapid amplification of 5Ј ends (Invitrogen) was performed according to the manufacturer's protocol, and the resultant cDNA clones were sequenced by 33 P-labeled dideoxynucleotide chain termination.
DNase I Hypersensitivity Analysis-Cultured cells were grown to near confluence, washed with phosphate-buffered saline, and collected by scraping. Cells were treated with lysis buffer (10 mM Tris, 10 mM NaCl, 3 mM MgCl 2 , and 0.5% Nonidet P-40), and recovered nuclei were exposed to DNase I (Roche Molecular Biochemicals) at concentrations of 0, 0.1, 0.3, 0.5, 0.7, and 1 g/ml at 37°C for 10 min. Partially digested DNA was recovered and extracted with phenol-chloroform, digested with XbaI overnight, and used in Southern analysis with a 32 P-labeled genomic DNA probe encoding bases Ϫ2132 to Ϫ1641 relative to the transcription start site. The hybridized filter was washed in 30 mM sodium chloride, 3 mM sodium citrate, 0.1% SDS at 65°C before autoradiography.
In Silico Analysis of the aif-1 Locus-To determine the likelihood of interaction of DNA in the vicinity of the aif-1 locus with the nuclear matrix, the genomic sequence (under GenBank TM accession number AF109719) from Ϫ6382 to ϩ6257 relative to the transcription start site was analyzed using the MAR-finder program (15) (Dr. Gautam B. Singh, www.futuresoft.org). Analysis of the 5Ј-flanking sequence for transcription factor binding sites was performed via the TESS server at the University of Pennsylvania (www.cbil.upenn.edu/tess) using the Transfac data base (11).
Genetic Reporter Studies-RAW264.7 and 3T3 cells (50,000/well) were transfected with FuGENE 6 reagent (Roche Molecular Biochemicals) on 24-well plates according to the manufacturer's protocol. In brief, up to 1 g/well total plasmid DNA was used in the experiments, including 0.1 g of a cytomegalovirus-␤-galactosidase reporter (CLON-TECH) used for normalization of transfection efficiency. Transfected cells were harvested 24 h later, and luciferase and ␤-galactosidase activities were determined. Luciferase activity for each well was corrected for variable transfection efficiency by dividing by the respective ␤-galactosidase activity. All transfections were performed in triplicate in at least three independent experiments.
Plasmids-To determine promoter activity of genomic sequences from the aif-1 locus, a 3.1-kb genomic DNA fragment was amplified by PCR using Klentaq (CLONTECH), forward primer (Ϫ2994 to Ϫ2970) 5Ј-GGTTAGAGGTCGGCTTGTTGGAA-3Ј, and reverse primer (ϩ88 to ϩ65) 5Ј-TTTGGCCCATGGCTCCTCAGACGC-3Ј, in which an NcoI site was introduced in-frame with the translation start site by a single base modification (G 3 C, underlined) of the genomic sequence. This PCR product was digested with NcoI, and the fragment from Ϫ902 to ϩ75 was cloned into the NcoI site in the pGL3 basic luciferase reporter (Promega). This construct was digested with HindIII and PvuII, and the 5Ј end was extended by ligation of the ϳ3.4-kb HindIII/PvuII genomic fragment (bases Ϫ4081 to Ϫ569) to generate the aif-1 Ϫ4082/ϩ75 luciferase reporter. Additional constructs were generated by digestion of this plasmid with restriction enzymes to yield aif-1 Ϫ1876/ϩ76, aif-1 Ϫ1286/ϩ75, and aif-1 Ϫ569/ϩ75 luciferase. The Ets mutant in aif-1 Ϫ902/ϩ75 luciferase was generated by site-directed mutagenesis (QuikChange, Stratagene) using primers with forward and reverse complements of the sequence 5Ј-TGGGGA-CAGCTGGTAGCTCTGTCTACTG-3Ј, in which the underlined bases replace the wild-type bases GAA to disrupt the Ets consensus binding site (AGGA). The STAT mutant was produced similarly using forward primer 5Ј-CTGTTCTGCAGAAGTGTCTCTTCAAAC-3Ј and reverse primer 5Ј-ACACTTCTGCAGAACAGAAAGTGAGA-3Ј; the underlined bases replace the wild-type bases CTCCA to disrupt the STAT consensus binding site (TTCCTCCAA). The IRF mutant was generated with forward primer 5Ј-CTACTGTCTCTTTCAGCTGCACTTTCT-GTTCCTCCA-3Ј and its reverse complement; the underlined bases replace the wild-type bases TCT to disrupt the IRF binding site (CTCTTTCAGTCTC). The aif-1 Ϫ902/Ϫ789 luciferase construct was generated by digestion of cloned aif-1 genomic DNA with NcoI, fill-in with Klenow enzyme, digestion with BglII, and ligation of the resultant fragment into the pGL3 promoter vector cut with SmaI and BglII. The aif-1 Ϫ902/ϩ1505 luciferase reporter was produced by the addition of an NcoI site between bases ϩ1505 and ϩ1510 by site-directed mutagenesis, digestion with NcoI, and ligation into NcoI-digested pGL3 basic. The Ets 3X luciferase concatamer was produced by annealing primer TGGTGGGGACAGGAAGTAGCTCTG-TGGTGGGGACAGGAAGTAG-CTCTGTGGTGGGGACAGGAAGTAGCTCTG to its reverse complement followed by digestion with Acc65I and XhoI to cut at restriction sites added to each end and ligation into the pGL3 promoter vector (Promega). Underlined bases denote the putative Ets binding sites. All constructs were confirmed by restriction digests and DNA sequencing. The following investigators generously provided plasmids: J. Leiden (Ets-1, Ets-2, Pu.1), X-Y. Fu (STAT1), D. Levy (IRF-9), K. Ozato (IRF-8), and B-Z. Levi (IRF-4).
Electromobility Shift Assays-Electromobility shift assays were performed as described (10) with modifications. The sense strand sequence of the DNA probe for the aif-1-Ets binding site (Ϫ894 to Ϫ880) is GGGACAGGAAGTAGC; the predicted consensus binding site is underlined. The forward sequence of the Pu.1 consensus probe is GGGCGCT-TGAGGAAGTATAAGAAT. The sense strand sequence of the probe for the aif-1 IRF site (Ϫ866 to Ϫ847) is TCTTTCAGTCTCACTTTCTG. These probes and their reverse complements were end-labeled with 32 P and annealed to one another. Typical binding reactions contained double-stranded DNA probe (20,000 cpm), 1 g of poly(dI-dC)⅐poly(dI-dC), 50 mM NaCl, 50 mM Tris HCl, pH 7.5, 1 mM MgCl 2 , 4% glycerol, 0.5 mM dithiothreitol, 0.5 mM EDTA, and 4 g of nuclear extract in a final volume of 20 l. The specificity of binding interactions was assessed by competition with a 100-fold excess of unlabeled double-stranded oligonucleotide of identical sequence or the Pu.1 consensus binding site sequence. The antibodies specific for Pu.1 and Sp1 were purchased from Santa Cruz Biotechnology.
Data Analysis-Quantitative results presented are representative of findings from at least three independent experiments. Comparisons among groups were made by factorial analysis of variance followed by a Bonferroni/Dunn post-hoc analysis. Statistical significance was accepted for a p value Ͻ0.05.

aif-1 mRNA Is Robustly Induced by IFN-␥ in Murine
Macrophages-To identify cell types with different levels of aif-1 gene transcription, we assessed aif-1 expression by Northern analysis in rat aortic smooth muscle cells, 3T3 fibroblasts, and RAW264.7 murine macrophage-like cells. The smooth muscle cell and fibroblast samples showed no signal (Fig. 1A); stimulation of smooth muscle cells with a panel of growth factors and cytokines likewise yielded no evidence of aif-1 expression (data not shown). In the macrophage samples, however, the cDNA probe readily identified a band with a calculated size of ϳ970 nucleotides. Because expression of aif-1 has been reported in conditions in which cells of monocyte/macrophage lineage are likely to be activated (2, 3, 5), we then assessed aif-1 mRNA expression in RAW264.7 cells exposed to a panel of cytokines and growth factors. As shown in Fig. 1B, most of these agents had a limited effect on aif-1 expression after 24 h of treatment. A substantial induction was seen, however, in cells treated with IFN-␥ (7.5-fold), whereas TGF-␤1 produced a more modest increase in aif-1 message (2.5-fold).
The aif-1 Gene Consists of 6 Exons Distributed Over ϳ1.6 kb-To determine the site of transcription initiation of the aif-1 gene, we performed 5Ј-rapid amplification of 5Ј ends, primer extension analysis, and RNase protection assays. The dominant transcription start site was most clearly identified by the primer extension analysis shown in Fig. 2A, with the thymidine (indicated in bold) in the antisense sequence corresponding to an adenine residue 78 bases upstream of the ATG encoding the initiation methionine on the sense strand. A fainter band indicated an alternative start site corresponding to a thymidine 2 bases downstream of the major site. RNase protection results were consistent with a major start site at or near 78 bases 5Ј to the translation start (data not shown), and sequences of the 5Ј rapid amplification of 5Ј ends isolates indicated transcripts starting 77-83 bases upstream of the translation start. Similar imprecision in the site of transcription initiation has been described in other macrophage-specific genes that do not have TATA-like sequences (12,13).
With the 5Ј end of the aif-1 transcript identified, it was possible to determine the genomic organization of the gene (Fig. 2B). Alignment of cDNA sequence to the corresponding genomic sequence indicated that the gene consists of 6 exons and 5 introns distributed over ϳ1.6 kb. The exons vary in size from 25 to 163 nucleotides, whereas introns range in size from 90 to 323 nt and conform to the GT/AG rule of splice donor/ acceptor recognition. The putative 5Ј EF-hand domain (3) is encoded by exons IV and V, whereas the 3Ј EF-hand domain, which may be ancestral, is entirely contained by exon V. Intron-exon structure involving EF-hand domains has been studied extensively; introns typically occur both within and between EF-hand domains, which argues against exon shuffling as an important factor in the evolution of EF-hand homolog proteins (14).

A DNase I Hypersensitive Site in Stimulated Macrophages
Lies ϳ0.9 kb Upstream of the Transcription Start Site-We then performed DNase I hypersensitivity analysis (8) of the aif-1 locus to identify potential changes in chromatin conformation linked to its characteristic expression pattern. We analyzed 3T3 cells, which have little or no aif-1 expression (Fig.  1A), RAW264.7 cells, which have moderate expression, and RAW264.7 cells stimulated with IFN-␥, which robustly express aif-1 mRNA (Fig. 1B). As shown in Fig. 3, A and B, the bands identified in the 3T3 DNA indicated that the entire aif-1 locus was not accessible to DNase I in our test conditions, as the size of band II was consistent with a hypersensitive site distal to the 3Ј end of the aif-1 gene. In contrast, analysis of the RAW264.7 cells showed a specific band (band III) that corresponds to the location of the aif-1 transcription start site. Moreover, after stimulation with IFN-␥, which increases aif-1 expression, a faster migrating band (band IV) appeared, consistent with the presence of an inducible hypersensitive site ϳ0.9 kb upstream of the transcription start site.
We also analyzed ϳ11 kb of sequence from the aif-1 locus in silico for potential association with the nuclear matrix using the MAR-finder program (Ref. 15, see also "Experimental Procedures"). The results of this analysis are presented in Fig. 3C. The exon-intron structure and DNase I hypersensitive sites described above are indicated as points of reference. This analysis provides an estimate of the probability of DNA association with the nuclear matrix based on the combinations of characteristic patterns in the sequence. The aif-1 locus, including the hypersensitive site near Ϫ0.9 kb and all 6 exons and 5 introns, coincides with an area of low probability of nuclear matrix association extending over ϳ4 kb, which is flanked on both sides by regions of high matrix association potential. The relatively low probability of nuclear matrix association of the aif-1 locus, including the regions of DNase I hypersensitivity (bands III and IV), may facilitate access by the RNA polymerase II and associated proteins for transcription, thus contributing to its inducibility.
Positive Transcriptional Activity Localizes to Sequences between Ϫ902 and Ϫ789 -Although matrix association probability indicated potential accessibility of sequences between Ϫ1.8 kb and ϩ2.5 kb, the DNase I hypersensitivity analysis suggested the regulatory importance of sequences near Ϫ0.9 kb in the aif-1 locus. We analyzed the 5Ј-flanking sequences for potential consensus binding sites for transcription factors that FIG. 1. Northern analysis of aif-1 expression in mouse cells. Total RNA (10 g) was subjected to Northern analysis with a 32 P-labeled murine aif-1 cDNA probe, and the nitrocellulose filter was washed to high stringency before autoradiography. The calculated size of the aif-1 mRNA is ϳ0.9 kb. Panel A, subconfluent rat aortic smooth muscle cells (SMC), 3T3 fibroblasts, or RAW267.4 macrophages were grown in 10% fetal calf serum before RNA extraction. Two independent samples were assessed for each cell type. Panel B, RAW267.4 cells cultured in 10% fetal calf serum were stimulated for 24 h with the indicated agents. Message intensity was normalized for loading variation by reference to intensity of the 28 S ribosome signal (lower panel) and expressed as fold increase relative to the mean of the two control sample intensities. Conditions tested are: ctl a, no stimulation, adherent to plastic; ctl b, no stimulation, non-adherent; IFN-␥, 300 units/ml; IL-1␤, 10 ng/ml; IL-4, 100 ng/ml; IL-10, 10 ng/ml; macrophage colony stimulating factor (MCSF), 50 IU/ml; platelet-derived growth factor BB (PDGF-BB), 10 ng/ml; transforming growth factor (TGF) ␤-1, 10 ng/ml; tumor necrosis factor-␣ (TNF), 100 u/ml; 12-O-tetradecanoylphorbol-13-acetate (phorbol ester), 5 nM. might contribute to its inducible expression in macrophages. As shown in Fig. 4, these sequences contain multiple sites that might interact with transcription factors associated with activation-associated and macrophage-specific gene expression, including an AP1 site near Ϫ920, a Pu.1 site near Ϫ888, a STAT site near Ϫ840, multiple Ets sites near Ϫ770, Ϫ498, and Ϫ210, and IRF-1 sites near Ϫ860 and Ϫ16. A consensus NFB binding site near Ϫ2110 lies outside the area of accessibility suggested by the nuclear matrix association probability analysis (Fig. 3C).
To evaluate the transcriptional function of sequences in the aif-1 locus, we performed transient transfection studies in RAW264.7 cells using different fragments of aif-1 genomic sequence driving expression of a luciferase reporter gene. The largest plasmid contained aif-1 sequences from Ϫ4082 to ϩ79 driving expression of the luciferase reporter; a series of 5Ј deletions was derived from this construct. As shown in Fig. 5A, constructs containing sequences upstream of Ϫ1876 showed activity similar to that of the pGL3 basic plasmid, which lacks a promoter. Progressive deletions from the 5Ј end revealed increasing promoter activity, however, with the greatest activity found in the Ϫ902/ϩ79 construct. A further 5Ј deletion construct, aif-1 Ϫ569/ϩ79 luciferase, lacked detectable promoter activity. Taking Ϫ902 as the 5Ј boundary of positive promoter activity, we then constructed a series of deletions from the 3Ј end, with aif-1 sequences driving expression of the minimal SV40 promoter in the pGL3 promoter vector. These reporters were all more active than the minimal promoter, but progressive removal of 3Ј sequences between Ϫ10 and Ϫ383 decreased activity, whereas successive deletion of sequences from Ϫ383 to Ϫ583 to Ϫ789 resulted in increased activity (Fig.  5B). This deletion series defined the 3Ј border of positive transcriptional activity at Ϫ789. Taken together, testing of the 5Ј and 3Ј deletion series indicated that most of the positive transcriptional activity in the aif-1 promoter lies between Ϫ902 and Ϫ789. Sequences upstream of this area have predominately repressive effects, whereas sequences between Ϫ789 and the transcription start site contribute both positive and negative effects to promoter activity.
In addition to sites upstream of the transcription start site, many genes contain important regulatory elements within introns. Our DNase I hypersensitivity analysis suggested that this was not the case with aif-1. To corroborate this, we also tested a reporter construct bearing the aif-1 locus from Ϫ902 to ϩ1505 containing the first 5 exons, all 5 introns, and the coding portion of exon 6. As expected (Fig. 5C), the activity of this plasmid was similar to that of aif-1 Ϫ902/ϩ79, suggesting that there is little intronic contribution to regulation of aif-1 expression.
Sequences between Ϫ902 and Ϫ789 Contain Most of the Macrophage-specific Activity of the aif-1 Promoter-Removal of repressive regulatory elements may lead to a loss of cell-type specificity of a promoter. With this in mind, we tested the reporter constructs in 3T3 fibroblasts and compared their activities to that obtained in RAW264.7 macrophages. As shown in Fig. 6A, the activity of the 5Ј deletion constructs in fibro- Total RNA (50 g) was used as template for reverse transcription reaction with a 32 P-labeled aif-1-specific primer (ϩ26 -43). The same primer (unlabeled) was used with 33 P-dideoxynucleotide termination to generate the accompanying (antisense) genomic sequence. RNAs tested were 3T3 (lane 1), mouse macrophage RAW264.7 (lane 2), and mouse macrophage RAW264.7 stimulated for 24 h with IFN-␥ (300 g/ml) (lane 3). Two thymidine residues corresponding to 5Ј end of the major extension products are indicated as capitals, with the dominant residue in bold. Panel B, exon-intron structure of the mouse aif-1 gene. The transcription start site 78 bases upstream of the initiation methionine codon is set as ϩ1 and is shown as a bold A. Exons are underlined and numbered I-VI, nucleotides are numbered on the left, and predicted amino acids are numbered on the right. Within the protein sequence, the putative EF-hand-like motifs (amino acids 58 -69 and 95-106) are indicated in bold. An asterisk marks the termination codon. blasts was close to or less than that of the promoterless pGL3 basic vector; in RAW264.7 cells, activity of the Ϫ902/ϩ79 construct was nearly 15-fold above base line. Sequences between Ϫ902 and Ϫ789 increased the reporter gene activity only slightly in fibroblasts, whereas in the macrophage cells these sequences conferred activity again close to 15-fold greater than the control vector. These findings suggest that the aif-1 promoter is regulated primarily by the interaction of positive transcriptional regulatory factors present in macrophages with sequences between Ϫ902 and Ϫ789. These promoter elements are not active in 3T3 cells, suggesting that these factors are not expressed in the fibroblasts or that their activity is subject to repression by additional factors present in this cell type.
The Ets Site at Ϫ891 to Ϫ883 Is Required for Basal and IFN-␥-inducible Promoter Activity in Macrophages and Binds Pu.1-These findings indicated the importance of cis-acting elements between Ϫ902 and Ϫ789 in the aif-1 promoter. Inspection of the sequence in this region revealed three potentially important elements, consensus binding sites for Ets (Ϫ891 to Ϫ883), IRF (Ϫ864 to Ϫ851), and STAT (-846 to Ϫ838). We anticipated that transcription factors binding these sites would work in a coordinate fashion to confer macrophagespecific and IFN-␥-inducible activity on the promoter. To assess their functional significance, we mutated these sequences in the context of the Ϫ902/ϩ79 reporter. As shown in Fig. 7A  (upper panel), mutation of either the Ets site or the IRF site markedly impaired the activity of the promoter. In addition, whereas the wild-type aif-1 Ϫ902/ϩ79 reporter was induced nearly 10-fold by IFN-␥ treatment, disruption of the Ets site or the IRF site completely eliminated this response. Contrary to our expectations, mutation of the consensus STAT site at Ϫ846 to Ϫ838 had only a slight effect on the overall promoter activity FIG. 3. DNase I hypersensitivity analysis and matrix association potential of the aif-1 locus. Panel A, cellular nuclei were exposed at 37°C for 10 min to increasing amounts of DNase I. DNA was extracted, digested overnight with XbaI, electrophoresed through 0.8% agarose, transferred to a nylon filter, hybridized with a 32 P-labeled DNA probe, and washed to high stringency. Panel B, schematic location of XbaI restriction and DNase I hypersensitive sites flanking the aif-1 locus. Hypersensitive sites corresponding to bands II, III, and IV and the position of the DNA probe (Ϫ2132 to Ϫ1641) are indicated. Panel C, matrix association potential of the aif-1 locus. Genomic sequence from Ϫ6382 to ϩ6257 relative to the aif-1 transcription start site was analyzed using the MAR-finder program (see "Experimental Procedures"). The aif-1 gene exon-intron structure (Fig. 2) and the location of DNase I hypersensitive sites (A and B) are depicted along the length of the sequence on the x axis. The predicted matrix association potential, on a scale of 0 -1.0, is plotted on the y axis. bp, base pair.
FIG. 4. Aif-1 5-flanking sequence, indicating cis-acting elements associated with macrophage-specific and inflammatory activation. Sequences between Ϫ2085 and Ϫ1000 are not shown, as indicated by the ellipsis. Consensus cis-acting elements are underlined, with putative interacting factors are denoted above the sequence. The transcription start site is in bold and labeled as ϩ1. and did not diminish its responsiveness to IFN-␥. Moreover, the Ets site alone was not sufficient to confer the response to IFN-␥, as activity of a concatamer of the Ets site actually decreased with IFN-␥ treatment (Fig. 7A, lower panel). Thus in the context of the aif-1 Ϫ902/ϩ79 promoter, the Ets and IRF sites are essential for both basal and IFN-␥-stimulated activity.
To identify the proteins binding to the Ets site at Ϫ891 to Ϫ883 and the IRF site at Ϫ864 to Ϫ851, we performed electromobility shift assays with nuclear proteins extracted from 3T3 and RAW264.7 cells. Although the radiolabeled Ets probe was only 15 bases in length, we found multiple bands in both cell types, suggesting that this sequence is capable of interacting with multiple nuclear protein species. A prominent band generated by macrophage extracts (Fig. 7B, lanes 3 and 4, arrow) and, to a much lower degree, by the 3T3 extract (lane 1, arrow) was specifically inhibited by an excess of both unlabeled identical competitor and by an unlabeled Pu.1 consensus competitor ( lanes 5 and 6, arrow). Incubation of the binding reactions with an antiserum specific for Pu.1 resulted in a supershifted band in RAW264.7 cells (lane 7, asterisk). An antiserum specific for the transcription factor Sp1, used here as a control, caused no change in nucleoprotein-probe migration. The identity of the supershifted band in RAW264.7 extracts with Pu.1 was confirmed by supershift analysis of 3T3 cells transfected with a Pu.1 cDNA expression plasmid (Fig. 7B, lanes 9 and 10).
The result of electromobility shift analysis with the probe derived from the aif-1 IRF site is shown in Fig. 7C. The slowest migrating band (band a) is specific, as indicated by its disappearance in the presence of unlabeled competing oligonucleotide, but is present in both 3T3 and RAW264.7 extracts. However, two faster migrating, specific bands are also seen; band b is present only in the RAW264.7 extracts (lanes 3 and 4), with slight increase in intensity with IFN-␥ treatment (lane 4), and band c is seen only in the fibroblast extract (lane 2). We tested several antibodies directed against candidate IRF family proteins (IRF-1, IRF-4, IRF-8, and IRF-9), but these failed to yield a supershifted band (data not shown).
Exogenous Pu.1 Is Sufficient to Activate the Promoter in Both Macrophage and Non-macrophage Cells-To test the ability of candidate transcription factors to interact with the aif-1 promoter, we performed transient assays with expression plasmids for Ets-1, Ets-2, STAT1, and Pu.1. Ets-1 and -2 activated the aif-1 Ϫ902/ϩ79 promoter modestly in RAW264.7 cells (ϳ1.5-and 2-fold above base line, respectively; Fig. 8A); in contrast, expression of Pu.1 caused a 6-fold increase in promoter activity. These Ets plasmids were then tested in 3T3 cells; as shown in Fig. 8B, Ets-1 and -2 had essentially no effect on the aif-1 Ϫ902/ϩ79 promoter in these cells, whereas Pu.1 alone was sufficient to activate the promoter, analogous to its effect in the macrophage cell line. Consistent with the limited effect of mutation of the consensus STAT site, co-transfection of a STAT1 expression plasmid did not activate the promoter significantly in either RAW264.7 or 3T3 cells. Along with the promoter mutation studies (Fig. 7), these findings further support a mechanism in which Pu.1 acts together with a factor binding to the nearby IRF site. We then assessed the ability of candidate IRF proteins to drive the aif-1 promoter alone and together with Pu.1. Consistent with the supershift analysis described above and in contrast to our findings with Pu.1, IRF-1, IRF-4, IRF-8, and IRF-9 both singly and in combination did not activate the promoter and had only a limited effect when added with Pu.1.

DISCUSSION
Although Aif-1 has been identified in several different experimental systems (2-4, 7), it is selectively expressed by cells of monocyte/macrophage lineage. aif-1 mRNA has been found by Northern analysis of samples from spleen (2, 3), testis (2, 3), and brain (2)(3)(4), with faint signals also present in lung and kidney (3). Because immunohistochemical analyses have typically localized Aif-1 expression to cells of the monocyte/macrophage lineage, its presence in these tissues is likely attributable to monocytes in residual blood or macrophages in tissue. One exception to this may occur in the testis, in which its expression by differentiating germ cells has been described (2). Because Aif-1 is selectively expressed in monocyte/macrophage cells and is specifically up-regulated in several disease models, characterization of the aif-1 gene, including transcriptional regulatory elements, may lead to better understanding of mo-lecular mechanisms controlling macrophage-specific gene expression in both basal and activated states.
To define the organization of the aif-1 gene, we first mapped its transcription start site. Consistent with other genes preferentially expressed in macrophages, such as macrosialin (12) and Fc␥RIIIA (13), we found some variability in the location of transcription initiation ( Fig. 2A), with the major start site occurring 78 bases upstream of the translation start site. Alignment of the aif-1 cDNA to genomic sequences from the aif-1 locus indicated a relatively compact gene, with 6 exons distributed over ϳ1.6 kb. This transcription start site, identified by primer extension and 5Ј-rapid amplification of 5Ј ends and supported by RNase protection (data not shown), applies to the transcript published as aif-1, but it is likely that additional 5Ј exons exist for some related transcripts, as BLAST searches using the 5Ј sequence unique to iba-1 cDNA show partial alignment with genomic sequences ϳ0.7, 3.4, and 3.6 kb upstream of the aif-1 transcription start site we have defined. 2 The assessment of matrix association potential (Fig. 3C) indicates that this entire aif-1 locus, including the upstream regulatory elements defined by DNase I hypersensitivity (Fig. 3, A and B) and reporter gene studies (Figs. 5-7), fits within a single uninterrupted region unlikely to associate with the nuclear matrix; additional 5Ј exons in the iba-1 transcript from near 3.4 and 3.6 kb upstream would fall between the peaks of matrix association potential at 2.3 and 4.1 kb upstream of the aif-1 start site. How this potential for nuclear matrix association relates to the tissue-specific and cytokine-mediated regulation of aif-1 expression is not precisely understood at present.
The 5Ј-flanking sequence of the aif-1 gene has multiple consensus binding sites for transcription factors associated with both macrophage-specific and inflammatory expression, including NFB, AP1, CCAAT/enhancer binding protein, IRF-1, and Ets. Identification of a DNase I hypersensitive site near Ϫ0.9 kb focused our attention on this region; the importance of this hypersensitive site was supported by 5Ј and 3Ј deletion analyses (Fig. 5), which localized positive transcriptional activity to sequences between Ϫ902 and Ϫ789. This region contained consensus binding sites for Ets (Ϫ883 to Ϫ891), IRF 2 N. Sibinga and M. Jain, unpublished observation. (Ϫ851 to Ϫ864), and STAT (Ϫ838 to Ϫ846) proteins; we anticipated that the Ets site would be important in controlling macrophage specificity and that the IRF or STAT sites would support IFN-␥ inducibility of the promoter.
Selective mutation of the Ets, IRF, and STAT sites between Ϫ883 and Ϫ838 in the aif-1 promoter demonstrated that the Ets and the IRF binding sites were required for both basal and IFN-␥-stimulated promoter function, whereas the STAT site was not important for either activity (Fig. 7). The interaction of the Ets family protein Pu.1 with the Ets site was supported by electromobility shift analysis (Fig. 7B), and co-transfection experiments (Fig. 8) indicated that Pu.1 and not Ets-1 or Ets-2 was sufficient to drive expression of the aif-1 Ϫ902/ϩ79 promoter in both macrophage and non-macrophage cell types. Pu.1 (16), initially characterized by virtue of its interaction with the purine-rich 5Ј-GAGGAA-3Ј site in the murine MHC class II I-Ab gene (17), has important regulatory functions in expression of genes in the lymphoid and myeloid lineages. When Pu.1 is disrupted by gene targeting, no mature lymphocytes or macrophages develop, and neutrophils that survive show incomplete and aberrant maturation (18). This maturation defect is cell-autonomous, because expression of mature macrophage markers can be rescued by reintroduction of Pu.1 into Pu.1 Ϫ/Ϫ hematopoietic precursors (19). Putative Pu.1 target genes expressed in macrophages include the colony stimulating factor 1 receptor (20), Fc␥RIIIA (13), Mac1/CD11b (21,24), and the macrophage scavenger receptor (22).
An intact IRF site was required, like the Ets site, for promoter activity in both basal and IFN-␥-stimulated conditions (Fig. 7A). Because Pu.1 activated the aif-1 promoter in both 3T3 fibroblasts and RAW264.7 macrophage cells, expression of the factor or factors binding to the IRF site is probably not restricted to the macrophage. Moreover, the Ets site in isolation was not sufficient (Fig. 7A, lower panel) to mediate the IFN-␥ responsiveness characteristic of the larger promoter constructs. Taken together, these results suggest that endogenous Pu.1 in macrophages and exogenous Pu.1 forcibly expressed in 3T3 cells interacts with more broadly expressed transcription factors binding to the nearby IRF site. Full activity of other Pu.1-dependent promoters has been found in several cases to depend on additional factors, including members of the AP1 (22), E2A, and IRF (23) families, which interact either directly or through formation of a DNA-dependent ternary complex with Pu.1.
The IRF site at bases Ϫ854 to Ϫ861 fits the consensus binding sites for IRF-1, IRF-8, and IRF-9. Electromobility shift analysis using a probe based on the IRF site sequence (Fig. 7C) identified three bands between the two cell types examined, one present in both 3T3 and RAW264.7 cells and one unique to each cell type. These band patterns are consistent with three different hypothetical mechanisms that could account for the aif-1 promoter regulation we have observed; they are 1) a negative regulatory activity present in both 3T3 and RAW264.7 cells (band a), which might be overcome specifically in macrophages by a positive activity restricted to macrophages and binding to a site not included in the IRF probe used here; 2) a positive regulatory binding activity restricted to RAW264.7 extracts that increases slightly in response to IFN-␥ treatment (band b); and 3), a negative regulatory binding activity present in 3T3 cells but not in RAW264.7 (band c). Pu.1 is typically a positive regulator of transcription and, with its restricted pattern of expression, could supply the positive activity required in the first of these mechanisms if band a in fact reflects the binding of a transcriptional repressor.
Of the candidate IRF proteins listed above, IRF-1 and IRF-9 are transcriptional activators and are expressed in many cell types (25,26). IRF-8, on the other hand, is typically a repressor of transcription, with expression restricted to cells of lymphoid and myeloid lineages (27). The related factor IRF-4 can function both as a repressor and activator, has been identified in myeloid as well as lymphoid cells, and has been shown to synergize with Pu.1 in activation of the IL-1␤ promoter (28). Thus, with the possible exception of IRF-4, which might act as a macrophage-restricted transcriptional activator along with Pu.1, these factors do not fit well with the mechanisms suggested by the electromobility shift results with the aif-1 IRF site probe.
We were not able to identify these IRF family members in cellular extracts that interacted with the site in electromobility supershift assays (data not shown). Although this finding might reflect technical limitations of the anti-IRF antibodies tested in electromobility shift assays or could stem from the requirement of some IRF proteins, including IRF-4 and IRF-8, for interaction with other proteins to bind DNA efficiently (23), it is consistent with the results of our functional assessments. As mentioned above, evaluation of expression plasmids for IRF proteins IRF-1, IRF-4, IRF-8, and IRF-9 by cotransfection with Pu.1 did not add significantly to the level of activation attained by Pu.1 alone.
Comparison of the aif-1 sequence encompassing the Pu.1 and IRF binding sites with previously described composite Pu.1/ IRF motifs (28) reveals differences that may be significant; the aif-1 sites are separated by ϳ21 bases, whereas the composite motifs have only 2 bases between the core elements. In addition, the orientation of the sites in the aif-1 sequence relative to one another differs from that in the composite elements. These characteristics of spacing and orientation may underlie the functional differences we have found in the responsiveness of the aif-1 sequences to added IRF proteins. The possibility that additional, perhaps novel or less well characterized members of the IRF family might be involved in the regulation of the aif-1 gene must also be considered. Further studies will be required to identify definitively the additional proteins that act together with Pu.1 to control transcription of the aif-1 gene by macrophages in inflammatory settings.