Isolation and Characterization of a Human STAT1 Gene Regulatory Element INDUCIBILITY BY INTERFERON (IFN) TYPES I AND II AND ROLE OF IFN REGULATORY FACTOR-1*

The transcription factor STAT1 plays a pivotal role in signal transduction of type I and II interferons (IFNs). STAT1 activation leads to changes in expression of key regulatory genes encoding caspases and cell cycle inhib-itors. Deficient STAT1 expression in human cancer cells and virally mediated inhibition of STAT1 function have been associated with cellular resistance to IFNs and mycobacterial infection in humans. Thus, given the relative importance of STAT1, we isolated and characterized a human STAT1 intronic enhancer region display-ing IFN-regulated activity. Functional analyses by transient expression identified a repressor region and type I and II IFN-inducible elements within the STAT1 enhancer sequence. A candidate IRF-E/GAS/IRF-E (IGI) sequence containing GAAANN nucleotide repeats was shown by gel shift assay to bind to IFN regulatory fac-tor-1 (IRF-1), but not to IFN-stimulated gene factor-3 (ISGF-3) or STAT1–3. An additional larger IGI-binding complex containing IRF-1 was identified. Mutation of the GAAANN repeats within the IGI DNA element elim-inated IRF-1 binding and the IFN-regulated activity of the STAT1 intronic enhancer region. Transfection of the IFN-resistant MM96 cell line to express increased levels of IRF-1 protein also sites primers. resulting constructs, pEF- STAT1 and pEF-IRF-1, purified by cesium chloride gradient centrifugation and used to transfect MM96 cells via electroporation. Stable clones were selected by culture in the presence of G418. Antiviral Assays— Antiviral assays were performed as described previously (7, 8).

The transcription factor STAT1 1 is activated by tyrosine phosphorylation mediated by JAK family kinases during cellu-lar responses to cytokine or growth factor signaling (reviewed in Ref. 1). STAT1 directly regulates expression of key proteins involved in controlling the cellular processes of growth arrest (2,3) by inducing expression of p21 WAF/CIP (4,5) and cell death via expression of caspases (6). Our work (7,8) and that of others (9 -11) revealed that STAT1 expression is deficient in a number of different human cancer cell types resistant to type I IFNs in antiproliferative and antiviral assays. STAT1 has also been shown to be the target of viral proteins inhibiting IFN signaling responses (12). In addition, STAT1 expression in tumor cells has been shown to be important for tumor elimination by immune surveillance mechanisms (13) as well as for mycobacterial immunity in humans (14). STAT1 is an integral component in a range of different transcription factor complexes. For example, the two most commonly identified are GAF and ISGF3, involved in the signal transduction pathways of type I and II IFNs, respectively (reviewed in Refs. 1 and 15). The GAF DNA-binding sequence (GAS) exhibits dyad symmetry across two half-sites, consistent with the binding of STAT1 proteins as a homodimer. Related GAS elements with between 2 and 4 intervening nucleotides bind transcription factors that are activated not only by IFNs, but also by numerous other cytokines and growth factors. Thus, GAS elements are capable of binding homo-and heterodimers of the STAT proteins STAT1, STAT3, STAT4, STAT5, and STAT6. A heterodimer of STAT1 and STAT2 can bind weakly to the GAS element, resulting in transactivation (16).
ISGF3 is recognized as the major multisubunit transcription factor activated in response to type I IFNs. ISGF3 comprises two components, ISGF3␥ or IRF-9, a single 48-kDa DNA-binding protein, and ISGF3␣, which consists of the STAT1␣ (91 kDa) or STAT1␤ (84 kDa) protein in a heterodimeric complex with STAT2 (113 kDa). Treatment of cells with IFNs can also result in association of homodimers of STAT1 or STAT2 in ISGF3. Complexes of STAT2 homodimers with IRF-9 may form, but have only weak DNA binding and weak transactivating potential. Type II IFN stimulation can result in formation of ISGF3 as a STAT1 homodimer⅐IRF-9 complex (17). Thus, a number of different STAT1 complexes have been identified to date.
Once formed, the ISGF3 complexes translocate to the nucleus and bind to the ISRE to activate ISG expression. The ISRE sequence is orientation-independent and contains direct repeats of either 5Ј-TTTC-3Ј (for example, see Ref. 18) or, more commonly, the complement 5Ј-GAAA-3Ј, and the repeats are usually separated by a spacing of 1-2 nucleotides (reviewed in Ref. 19). The ISG15 ISRE with sequence 5Ј-CTCGGGAAAGG-GAAACCGAAACTGAAGCC-3Ј binds to two different IFN-inducible transcription factors and has been studied extensively. Not only is the core sequence (the nonamer underlined in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequences reported in this paper have been submitted to the GenBank TM /EBI Data Bank under accession numbers AF182310 -AF182313.
Here we report the mapping of the human STAT1 gene from the Centre d'Etude du Polymorphisme Humain human YAC library using a PCR-based screening approach, isolation and characterization of the 5Ј-genomic region, and identification of a 5Ј-transcriptional start site. Putative DNA-binding sites for transcription factors were identified within the intron 1/exon 2 boundary sequence. Transient expression assays with luciferase reporter constructs and gel mobility shift assays confirmed the presence of an unusual ISRE (IRF-E/GAS/IRF-E) binding a high molecular mass IRF-1 complex that included CBP and was induced by stimulation with either type I or II IFN. Interestingly, we have not found any binding of ISGF3 components to this ISRE. We propose that the IRF-1 and STAT1 genes and their proteins form an intracellular amplifier circuit regulating cellular responsiveness to the IFNs.

MATERIALS AND METHODS
Cell Culture and Stimulation-Melanoma cell line SK-MEL-28 (American type Culture Collection, Manassas, VA) was grown in RPMI 1640 medium supplemented with 10% inactivated fetal calf serum. Human MCF-7 breast cancer and U293T embryonic kidney cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% inactivated fetal calf serum. In all experiments, cells were stimulated with either 1000 IU/ml IFN-␣2a or 1000 IU/ml IFN-␥ (Hoffmann-La Roche, Basel, Switzerland) for the indicated time periods. For IFN-␥ priming, cells were pretreated with 1000 IU/ml IFN-␥ for 16 h prior to stimulation with IFN-␣2.
YAC Clones, DNA Purification, and Primers Used for Detecting the STAT1 Gene-YAC clones neighboring D2S118 or AFM066xc1 sequence tagged sites (STS), including 737c9, 755b5, 764F5, 805c1, 809e2, 820b12, 821e5, 855g4, 855h9, 984a5, 894a6, and 894a11, were obtained from Dr. Simon Foote (Walter and Eliza Hall Institute for Medical Research, Parkville, Victoria). A single colony of each YAC strain was expanded by growth in 10-ml cultures as described (20), and the DNA was purified as described, washed with 80% ethanol before it was dried, and resuspended in 50 l of MilliQ water. For rapid preparation of yeast for analysis by PCR, a yeast colony (0.5-2 mm) taken from a plate was lysed by addition of 2-10 l of incubation solution (100 mM sodium phosphate (pH 7.4) and 1.2 M sorbitol) containing zymolyase 20T (2.5 mg/1 ml of incubation solution) and incubated at 37°C for 5 min. The digested product was ready to be used in the PCR or was stored at Ϫ20°C for future use. The primers used were as follows: STAT1yac3Јa (18-mer), 5Ј-CATCTTCTCTGGCGACAG-3Ј, corresponding to nucleotides 2432-2449 of STAT1 cDNA sequence (GenBank TM /EBI accession number M97935); and STAT1yac3Јb (20-mer), 5Ј-CAGTAAGATGCAT-GATGCCC-3Ј, corresponding to nucleotides 2599 -2618 of STAT1 cDNA sequence.
Bubble PCR and Primers Used to Isolate the STAT1 5Ј-Genomic Region-The procedures for bubble PCR were as described previously (21) and formed the basis for the methods used here. The oligonucleotide sequences used for the construction of bubble anchors/vectorettes were as follows: BubbleVec Top primer (55-mer), 5Ј-GCTCTCCCTTCT-GCGGCCGCAGTTCGTCAACATAGCATTTCTGTCCTCTCCTTCG-3Ј; and BubbleVec Bottom primer (60-mer), 5Ј-AATTCGAAGGAGAGGA-CGCTGTCTGTCGAAGGTAAACGGACGAGAGAAGGGAGAGCTGCA-3Ј. Bubble anchors were annealed and ligated to EcoRI-and PstIdigested DNA from either the 805c1 or 809e2 YAC clone via the overhanging ends of the anchors/vectorettes. Prior to PCR, ligated products were extracted from excess bubble anchors/vectorettes by passage through a QIAQuick PCR purification kit (QIAGEN Inc.) as out-lined by the manufacturer. The final products were used as templates for bubble PCR with the "hot start" procedure and platinum Taq polymerase (Invitrogen) in a total volume of 50 l, consisting of 5 ng of anchored DNA template and 1 M primers. One of the primers used was either BubbleprimerIntE or BubbleprimerIntP (described below), which would prime from the anchor/vectorette at the 5Ј-end of the target DNA; the other primer was based on the nucleotide sequence of the STAT1 cDNA (GenBank TM /EBI accession number M97935). PCR cycling parameters were as follows: 94°C for 3 min, followed by 35 cycles of 94°C for 15 s, 63°C for 10 s, and 72°C for 3 min. The product was gel-purified before cloning into the pGEMT-Easy vector used for DNA sequencing reactions (Big Dye sequencing kit, PerkinElmer Life Sciences). The primers used were as follows: BubbleprimerIntE (30-mer), 5Ј-GCGGC-CGCAGTTCGTCAACATAGCATTTCT-3Ј; BubbleprimerIntP (27-mer), 5Ј-GCTGTCTGTCGAAGGTAAACGGACGAG-3Ј. In addition, three primers were prepared as complementary sequences corresponding to the indicated regions from the STAT1 cDNA sequence (GenBank TM / EBI accession number M97935) as follows: STAT1 prom primer (30 mer), 5Ј-GGAATTCCGCTGCTGAAGTTCGTACCACTG-3Ј, corresponding to nucleotides 203-224; STAT1 upstream primer, 5Ј-CTTTCCGG-CGCAGAGTCTGC-3Ј, corresponding to nucleotides 26 -46; and STAT1 upRev primer, 5Ј-CTGCGGAGGGGCTCGGCGAG-3Ј, corresponding to nucleotides 10 -29.
RNA Ligase-mediated Rapid Amplification of cDNA Ends (RLM-RACE)-Messenger RNA was isolated from SK-MEL-28 cells treated with IFN-␥ (1000 IU/ml, incubated for 16 h) using the FastTrack 2.0 mRNA isolation kit (Invitrogen). To obtain the full-length 5Ј-end of the human STAT1 cDNA, full-length RLM-RACE was performed using the GeneRacer kit (Invitrogen) according to the protocols recommended by the manufacturer. The GeneRacer kit ensures that only full-length transcripts are amplified by prior elimination of truncated transcripts. The STAT1 gene-specific primer (STAT1prom) was used to synthesize the first-strand cDNA template. Amplified fragments resulting from RLM-RACE were cloned into pCR4Blunt-TOPO ® (Invitrogen), and the DNA sequence was determined using the DYEnamic ET terminator cycle sequencing kit (Amersham Biosciences).
Luciferase Constructs and Site-directed Mutagenesis-The pGL3-Basic luciferase reporter vector (Promega) encoding a modified firefly luciferase was used to prepare STAT1 intronic enhancer reporter constructs. The primer Rvprimer3 (5Ј-CTAGCAAAATAGGCTGTCCC-3Ј) was used for DNA sequencing across the regions cloned into pGL3basic to confirm that the sequences of the constructs were correct. The constructs prepared included the following. 1) The full-length (FL) construct contained the full-length EcoRI fragment of the STAT1900prom region obtained from pGEMT-Easy. 2) The 3Ј-luc (where luc is luciferase) vector contained the SacI/XhoI 3Ј-half of the STAT1 intronic enhancer region isolated from pGEMT-Easy (from bp ϩ508 to ϩ1223 relative to the STAT1 transcriptional start site).
3) The 5Ј-luc vector in which the 5Ј-half of the STAT1 intronic enhancer (from bp ϩ3 to ϩ498) was derived from the full-length clone by PCR with the p91-ISREup and STAT1upstream primers, and the product was cloned into pGEMT-Easy, excised with PstI/NotI, ligated into pCR-Script SK ϩ (Stratagene), re-excised with KpnI/SacI, and cloned into pGL3Luc. 4) The 5Ј⌬IGI-luc construct was made by inserting a PCR-derived fragment of the 5Ј-luc construct between the KpnI and SacI sites of pGL3basic. The PCR primers used were as follows: forward primer, GAC GGT ACC GCA CGA CTG GCA AGG ACA ATC; and reverse primer, GCA CTC GAG GAA TTC GAT TCT TTC CGG CGC. 5) For IGI-luc, p91-ISREup and its complement p91-ISREdown were annealed at 50°C and cooled down slowly to room temperature to allow double-stranded annealing before cloning into KpnI/BglII-cut pGL3Luc. The sequence of the IGI-luc construct was confirmed using the forward primer.
Mutation of the GAAANN sites within the IGI element of the pGL3-FL construct was carried out using the QuikChange site-directed mutagenesis kit (Stratagene) as described by manufacturer. Briefly, the pGL3-STAT1prom FL construct was subjected to PCR with primers hSTATmut/sense (5Ј-CCATCAGTCACGTTTGATTCTTTTAGCAGAA-AGTTGGATTTGCTGTATGCC-3Ј) and hSTATmut/anti (5Ј-GGCATAC-AGCAAATCCAACTTTCTGCTAAAAGAATCAAACGTGACTGATGG-3Ј) for site-directed mutagenesis. After amplification, parental DNA was removed by digestion with DpnI, and the vector DNA containing the mutant IGI element was transformed into rubidium chloridetreated, competent Escherichia coli DH5␣ cells. The DNA sequences of the mutant clones were verified using the DYEnamic ET terminator cycle sequencing kit.
Cell Transfection and Reporter Assays-Cells (2 ϫ 10 5 ) growing in 35-mm dishes were transfected with 1 g of plasmid DNA/dish using purified pSV2-␤gal (Promega) and/or pGL3basic, 5Ј-luc, IGI-luc, 3Ј-luc, FL-luc, 5Ј⌬IGI-luc, FL-IGImt-luc, or pEF-luc DNA. Either FuGENE 6 (Roche Molecular Biochemicals, Mannheim, Germany) or Transfast reagent (Promega) was used according to the manufacturers' protocols. At 30 -36 h after transfection, cells were either left untreated or stimulated with 1000 IU/ml IFN-␥ or IFN-␣2 and incubated for 8 -24 h before harvesting, and extracts were prepared for luciferase and ␤-galactosidase assays using standard protocols (Promega). Light units were measured in a Top Count microplate scintillation counter (Canberra Packard), recording four readings over 2 min and averaging these to measure light units/min. Each transfection assay was performed in triplicate and repeated two to three times; the values (light units/min) were averaged; and S.E. values were calculated.
RNA Preparation and Northern Analysis-Total RNA was prepared using Trizol (Invitrogen) according to the manufacturer's instructions. For Northern blotting, 20 g of total RNA was separated on 1.5% formaldehyde gels and transferred to Hybond Plus membranes (Amersham Biosciences). The STAT1 probe for Northern blotting was prepared by PCR using the mouse Stat1 cDNA as a template. The primers used to obtain the 177-bp fragment were as follows: forward primer, CCGATG GAG CTT GAC GAC CCT AAG CGA AC; and reverse primer, CTG TGC TCA TCA TAC TGT CAA ATT CGG CC. The STAT1 probe was labeled with [␣-32 P] dCTP and hybridized to the membrane-bound RNA overnight at 65°C in Church's buffer (0.158 M NaH 2 PO 4 , 0.342 M Na 2 HPO 4 , 1 mM EDTA, 1% bovine serum albumin, and 7% SDS). Following stringent washing, membranes were exposed to film or Phos-phorImager screens. Transfers were stripped and reprobed with the GAPDH probe (pTRI-GAPDH, Ambion Inc.) to assay for GAPDH expression.
Protein-DNA Cross-linking in Vitro and Western Blotting-Streptavidin-agarose beads (NeutrAvidin TM -agarose, Pierce) were saturated with 5Ј-biotinylated IGI oligonucleotides at 4°C for 2 h. The beads were then washed with HEM buffer and incubated with extracts from untreated or IFN-␥-treated MCF-7 cells at 4°C for 1 h. The beads were washed five times with HEM buffer, resuspended in SDS-PAGE loading buffer, and loaded onto 10% SDS-polyacrylamide gels. Extracts from MCF-7 cells were quantitated, and 20 g of each sample was separated on 10% SDS-polyacrylamide gels. Proteins were transferred to polyvinylidene difluoride membranes and analyzed by Western blotting with either anti-CBP or anti-IRF-1 antibody as indicated.
The full-length IRF-1 expression vector was generated by PCR using the IRF-1 5Ј-primer (5Ј-AAGGAAAAAAGCGGCCGCCACCATGGCCC-CATCACTCGGATGCGCATGAGACCC-3Ј) and the IRF-1 3Ј-primer (5Ј-GCTCTAGACTACGGTGCACAGGGAATGGCC-3Ј) from the IRF-1-pBluescript II SK plasmid as a template (kindly provided by Dr. Richard Pine). The PCR product was cloned into the pEFMCIneopAN9 expression vector (23) via NotI and XbaI restriction enzyme sites introduced by the above primers. The resulting constructs, pEF-STAT1 and pEF-IRF-1, were purified by cesium chloride gradient centrifugation and used to transfect MM96 cells via electroporation. Stable clones were selected by culture in the presence of G418.

RESULTS
Mapping of the STAT1 Gene from the CEPH Human YAC Library-When this work commenced, the DNA sequence of human chromosome 2 was incomplete, and the 5Ј-genomic sequence of the human STAT1 gene was unknown. The STAT4 and STAT1 genes had been located between q32.2 to q32.3 on human chromosome 2 (24,25). The CEPH-Genethon Database 2 revealed STAT4 located at marker D2S118 or STS AFM066xc1. YACs from the CEPH human contig library (26) and overlapping or neighboring D2S118 and STS AFM066xc1 were selected and cultured, and DNA was obtained from each of clones 737c4, 755b5, 764f5, 805c1, 809e2, 820b12, 821e5, 855g4, 855h9, 894a5, 894a6, and 894a11. PCR was carried out on YAC DNA samples using primers STAT1yac3Јa and STAT1yac3Јb. DNA products predicted from the STAT1 cDNA sequence (27) of 187 bp were obtained from clone 805c1 and 809e2 YAC DNAs as templates (Fig. 1A). DNA from clones 805c1 and 809e2 was purified, digested with either PstI or EcoRI (selected because these are infrequent sites present every 4 -6 kb), and analyzed by Southern blotting with an ␣-32 P-labeled STAT1 cDNA fragment as a probe. Positive bands were identified, confirming the presence of the STAT1 gene (data not shown).
Use of Bubble PCR to Clone the Human STAT1 5Ј-Genomic Structure-We used bubble PCR (21) of PstI-and EcoRI-digested YAC DNA to obtain specific 5Ј-fragments of the STAT1 gene, and GenBank TM /EBI accession number M97935 provided the initial 5Ј-cDNA sequence information used as the basis for primer design. Initially, we selected the nucleotide STAT1prom ( Fig. 2) adjacent to the ATG initiation codon of STAT1 to use with either BubbleprimerIntE or BubbleprimerIntP as the primer set. Specific PCR products were obtained whether template 805c1 or 809e2 YAC DNA digested with either EcoRI or PstI was used (Fig. 1B). The bubble PCR fragments cloned were called PromIcl2 (2.2 kb, obtained with EcoRI) and PromIIcl3 (1.5 kb, obtained with PstI). The resulting DNA sequences determined from both fragments were aligned and overlapped at the 3Ј-end such that the larger fragment (PromIcl2) extended ϳ700 bases more 5Ј than PromIIcl3 (Fig. 2). Analysis of the sequence adjacent to the STAT1prom primer confirmed the intron/exon boundary previously identified (GenBank TM /EBI accession number U18666) (Fig. 2B). Attempts to use primers designed to be near the 5Ј-end of the published STAT1 cDNA nucleotide sequence (accession number M97935) (27) for bubble PCR were unsuccessful, and no specific products resulted from either the PstI-or EcoRI-digested 809e2 DNA used as a template (data not shown). However, using a primer farther in from the 5Ј-end (STAT1upstream, nucleotides 26 -45 of the STAT1 cDNA sequence) together with the BubbleprimerIntE primer for bubble PCR produced a specific PCR product of ϳ900 bp, STAT1900prom (Fig. 1C). No specific PCR product could be detected when the PstI-digested 805c1 anchor was used as the target template. DNA sequencing of STAT1900prom revealed that it was 1223 bases long and that the 3Ј-end overlapped with the 5Ј-end of the published cDNA sequence (accession number M97935) (see Fig. 4), but contained marked differences to the M97935 cDNA sequence. Thus, the 12 nucleotides at the 5Ј-end of the published cDNA sequence (27) are not present in the genomic sequence. In addition, our sequencing also revealed that the next 29 nucleotides in the published cDNA sequence are inverted in the reverse direction.
Characterization of the 5Ј-Transcriptional Start Site in the STAT1 Gene-To resolve the problems identified above with the published cDNA sequence as well as to determine the 5Ј-transcriptional start site of the human STAT1 cDNA, RLM-RACE was undertaken. A product of 225 bp was obtained (data not shown), which was cloned and sequenced. The start site was identified as shown in Fig. 2, and the region of 5Ј-cDNA sequence overlapping with the 5Ј-DNA sequence from STAT1900prom was found to be identical (Fig. 2). Comparative analysis of the resulting cDNA and genomic sequences indicated the presence of two exons in the STAT1900prom DNA fragment (exons 1 and exon 2) (Fig. 2).
Human STAT1 Genomic Intron 1 Contains an IFN-regulated Enhancer Element-MatInspector Version 2.2 3 analysis of the STAT1900prom sequence revealed the presence of IRF-E elements, two GAS elements, and a retinoic acid response element (RARE, highlighted in Fig. 2B). One unusual feature of the sequence was the presence of one of the GAS sites in between two IRF-Es that each contained GAAANN repeats (boxed sequence in Fig. 2B). Previous studies have identified the presence of intronic enhancers within several IFN-regulated genes (28,29). Given these observations, vectors were constructed using different regions of the STAT1900prom fragment subcloned into the basal luciferase reporter plasmid, pGL3basic (Fig. 3A). The constructs were transiently transfected into cells that were then induced with either IFN-␥ or IFN-␣ and analyzed for the levels of induced luciferase activity (Fig. 3, B-D). Constructs containing either the whole region or either halves of the human STAT1900prom 5Ј-genomic region were examined in three different human cell lines, MCF-7, U293T, and SK-MEL-28. The results were very consistent (Fig. 3, B-D). In two of the cell lines, the 3Ј-portion (3Ј-luc) produced a dramatic reduction in the level of constitutive luciferase activity, ϳ10fold lower than the control levels resulting from the basal pGL3basic vector alone, and showed no effect of IFN treatment (Fig. 3, B and C). This result indicates the presence of a repressor element in the 5Ј-end of the STAT1 genomic intron 2 region. By contrast, in the absence of added IFNs, the 5Јportion of STAT1900prom (5Ј-luc) showed constitutive luciferase activity, ϳ5-10-fold above the pGL3basic background levels, and a further 2-3-fold increase in activity when induced by either IFN-␥ or IFN-␣ (Fig. 3, B-D). However, the full-length STAT1900prom construct, in the absence of IFNs, produced a similar low level of activity compared with the basal pGL3basic vector, but was induced 5-10-fold after treating cells with either IFN-␥ or IFN-␣ for 16 h. The kinetics of induction were similar when type I or II IFN was used to induce the full-length or 3Ј-luc constructs such that the luciferase activity peaked between 16 and 24 h after IFN induction.
Site-directed mutagenesis of the full-length STAT1900prom construct was used to specifically modify the GAAANN repeat 3 Available at www.gsf.de/cgi-bin/matsearch2.pl. FIG. 1. A, YACs containing the STAT1 gene in regions of human chromosome 2 identified by PCR. DNA was obtained from each of the YAC clones 737c4, 755b5, 764f5, 805c1, 809e2, 820b12, 821e5, 855g4, 855h9, 894a5, 894a6, and 894a11 using the zymolyase 20T rapid lysis method. The STAT1yac3Јa and STAT1yac3Јb primers (corresponding to nucleotides 2432-2449 and 2599 -2618, respectively, of STAT1 cDNA sequence (GenBank TM /EBI accession number M97935)) were used for PCR, and the products were analyzed on a 1.5% agarose gel. M, pUC19/ HpaI markers. B, use of bubble PCR to isolate the human STAT1 intron 2/exon 3 boundary. EcoRI-and PstI-digested 805c1 and 809e2 YAC DNAs were ligated to bubble anchors as indicated before PCR. Products were analyzed on a 1% agarose gel. M, DNA EcoRI/HindIII markers. Products were designated PromIcl2 and PromIIcl3 as indicated by the arrows. C, isolation of the human STAT1 5Ј-genomic region by bubble PCR. PstI-and EcoRI-digested 805c1 YAC DNA was ligated to bubble anchors before PCR with the STAT1upstream primer, and the products were analyzed on a 1% agarose gel. M, DNA EcoRI/HindIII markers. sites contained within the IGI element. When the full-length reporter gene construct with altered GAAANN sites was assayed, the results (Fig. 3B) revealed a complete loss of the responsiveness of the region to IFN-␥ and IFN-␤.
The IGI Element within the Human STAT1 Intronic Enhancer Binds a Novel IFN-inducible Complex That Contains the IRF-1 Protein-The above results indicated the presence of an IFN-responsive element contained within the STAT1900prom region, and the IGI element was implicated as the most likely site of IFN regulation. Therefore, EMSA was carried out using double-stranded oligonucleotide sequences encompassing the IGI element as probes. Extracts prepared from the three different human cell lines (MCF-7, U293T, and SK-MEL-28) were used in EMSA reactions, the latter cell line having been extensively characterized in previous studies as highly sensitive to both type I and II IFNs (7,8). In control EMSA reactions (data not shown), the use of the ISG15 ISRE and guanylate-binding protein GAS oligonucleotides revealed potent induction of ISGF3 and GAF activities in extracts from SK-MEL-28 cells treated with IFN-␣ and IFN-␥, respectively. In addition, ISGF3 was successfully supershifted by addition of anti-IRF-9 antibody, and anti-IRF-1 antibody supershifted the IRF-1 bands obtained with extracts from cells treated with either IFN-␥ or IFN-␣ (data not shown), confirming the specificity of these antibodies for supershift experiments.
With samples from all three cell lines, when EMSA was carried out using the ␥-32 P-radiolabeled IGI DNA element as the probe (Fig. 4), a single prominent IFN-induced band was observed, particularly after incubation with extracts from IFN-␥-or IFN-␣-treated cells. Low levels of the IGI complex were detected in some preparations of untreated cells (Fig. 4). However, the amount of IGI complex detected was greatly increased after IFN treatment and was especially noticeable in EMSA reactions of cell samples that had been previously primed with IFN-␥ for 16 h before IFN-␣ activation for 30 min (data not shown). The levels of the novel IGI-binding complex obtained from IFN-␥-treated MCF-7 cells were higher in the 30-min than in the 60-min treated samples (Fig. 4A). The IGI complex exhibited slower mobility in EMSA than the STAT1 homodimeric complex bound to the GAS element from the human IRF-1 gene (Fig. 4A). A smaller oligonucleotide probe containing the GAS element within the IGI sequence motif of STAT1900prom (GAA ACT TTC TGC GAA AAG) was unable to bind the IFN-␥-induced STAT1 homodimer (Fig. 4A). In addition, recombinant STAT1 protein was assessed for its ability to bind to this IGI GAS element. Binding of the recombinant STAT1 dimer to the IGI GAS sequence was detected (Fig. 5A), albeit at low levels compared with levels of recombinant STAT1 protein binding to the GAS site from the human IRF-1 gene promoter (cf. Fig. 4A).
A significant fraction of the IGI complex was supershifted upon addition of anti-IRF-1 antibody (Fig. 4, B-D). However, addition of antibodies against STAT1, STAT2, STAT3, IRF-2, or IRF-9 (Fig. 4, B-D) or IRF-3 (data not shown) did not significantly affect the resulting levels of IGI complex formation. These results indicate that the IGI complex contains IRF-1 protein, but does not contain significant levels of any of the other IFN-activated transcription factors.
Mutations in the Human STAT1 Intronic Enhancer IGI Element Affect Its Ability to Form Protein Complexes-To determine which of the sites in the IGI element are important for binding to protein complexes, EMSA was carried out with ␥-32 P-radiolabeled oligonucleotides containing different mutations in the DNA sequences of the IGI enhancer elements. When mutations were made individually to either the 5Ј-or 3Ј-IRF-E GAAANN repeat elements in the IGI sequence, they resulted in a reduced level of IGI complex formation (Fig. 5B). Mutation of the IGI GAS domain (TTCTGCGAA) to TT- GCTAGCGAA (p91-ISRE⌬gas1) (Fig. 5C) did not affect the formation of the IFN-activated IRF-1-containing IGI complex. However, mutations to the GAAAN repeats that included a mutation of the GAAANN site overlapping the IGI GAS 3Ј-end (TTCTGCGAAAGT to TTCTGCTAAGT) resulted in significant reductions in IGI complex formation. Thus, IGI mutant 1 (p91-ISREmut1 and its complement p91-ISREmut2) contained mutations in three GAAAN repeats, whereas IGI mutant 2 (p91-ISREmut3 and its complement p91-ISREmut4) contained mutations in only the two 3Ј-GAAAN repeat sequences. The 5Ј-GAAAN sequence was mutated to CCAAN, the middle sequence to TAAAN, and the 3Ј-sequence to TCAAN. With either the IGI mutant 1 or 2 oligonucleotides, the EMSA binding of the IRF-1-containing IGI complex was no longer detected (Fig.  5C), indicating that the IGI 3Ј-GAAAN sequences are critical for IRF-1 binding. No gel shift bands with comparable mobility to ISGF3 or STAT1-3 dimeric complexes were detected in any of the assays with any of the IGI probes examined.
Induction of IRF-1 Expression Alone Does Not Result in DNA Binding of the IGI Complex or Transcription of the Human STAT1 Gene-IRF-1 gene expression is regulated by activation of STAT1 (30), and the TNF-␣ and IFN-␥ signaling pathways converge to synergistically increase expression of IRF-1 via a composite GAS/NFB promoter element (31,32). Thus, we FIG. 3. Functional analysis of constitutive and IFN-induced human STAT1 intronic enhancer activity. A, schematic diagram depicting the STAT1 promoter fragments cloned into the pGL3basic luciferase vector. The human STAT1 intron 1 region contains IFNinducible enhancer activity. The human STAT1 5Ј-genomic constructs analyzed included the full-length construct (exon 1 (e1)/intron 1/exon 2 (e2)/intron 2 5Ј-region), the 5Ј-end and 3Ј-end regions, the IGI element alone, the 5Ј-end with the IGI element deleted, and the full-length construct containing a mutant IGI element. Each DNA fragment was isolated by PCR, cloned into pGL3basic, and confirmed by DNA sequencing. B-C, mean total values (ϮS.E.) of luciferase activity (light units/min) normalized from sets of results from two to three repeated experiments. The cells indicated were transfected with reporter constructs as indicated and were treated or not with IFNs. Analyses were carried out with U293T (B) or SK-MEL-28 cells (C) transfected under standardized conditions using 1 g of purified plasmid DNA/reaction. In D, MCF-7 cells were transfected with the pSV2-␤gal construct and the empty vector (pGL3basic), pEF-luc (positive control), or the 5Ј-luc, 3Ј-luc, or IGI-luc construct. Transfected cells were split 36 h posttransfection and either left untreated or treated with IFN-␥ for 8 h. Extracts from untreated and IFN-␥-treated and transfected cells were then analyzed for ␤-galactosidase and luciferase activities according to standard protocols. Luciferase activity units in D were normalized against ␤-galactosidase activity for each sample. RARE, retinoic acid response element. FIG. 4. A, the IGI motif of the human STAT1 intronic enhancer binds a novel IFN-inducible complex. MCF-7 cells growing in Dulbecco's modified Eagle's medium ϩ 10% fetal calf serum were treated with IFN-␥ for 30 or 60 min before preparing whole cell extracts for EMSA. Oligonucleotide probes corresponding to the 3Ј-GAS site from the STAT1900prom sequence (TGATTCCATGAACAT and its complement), the IGI motif from the human STAT1 (hSTAT1) intron 1/exon 2 boundary (CAG CAA ATG AAA CTT TCT GCG AAA AGA AGA AAA CGT and its complement), and the GAS element from the human IRF-1 (hIRF-1) gene promoter were analyzed. Binding of recombinant STAT1 (rSTAT1) was also examined by adding 500 ng to the EMSA reaction. DNA-bound complexes were resolved on 6% polyacrylamide gels and detected by autoradiography. B-D, the novel IGI-binding complex contains IRF-1. Cell extracts from MCF-7 cells (B) as described for A or from SK-MEL-28 (C) or U293T cells (D) treated with IFN-␥ or IFN-␣2a for 30 min were used in EMSA reactions. Whereas the shorter IGI element of STAT1900prom was used in A and B, the slightly longer IGI element of STAT1900prom (p91-ISREup/down annealed together) was used in C and D. Supershift analysis of the novel IGI-binding complex was carried out by adding 1 g of the indicated antibodies (Ab) to the EMSA reaction before addition of the radiolabeled IGI probe (p91-ISREup/ down oligonucleotides).
compared the IFNs and TNF-␣ for their ability to induce increased levels of IRF-1 expression and whether the increased expression was associated with the increased formation of the IGI DNA-binding complex by these cytokines. MCF-7 cells were stimulated with IFN-␣, IFN-␥, or TNF-␣ either alone or in different combinations for 8 h to compare their effects on IRF-1 and STAT1 expression as well as activation of IGI complex binding. IFN-␥ and TNF-␣ both increased expression of the IRF-1 protein (Fig. 6). However, only IFN-␣-or IFN-␥-treated MCF-7 cells showed increased expression of STAT1 mRNA after 8 h, particularly in the IFN-␥-treated samples. In addition, activation of the IGI-binding complex was only detected in MCF-7 cells treated with IFN-␥. Thus, from these cellular studies, activation of the IGI DNA-binding complex was associated with increases in STAT1 mRNA expression, supporting a role for the IGI complex in activating cellular STAT1 gene expression. However, the presence of high IRF-1 protein levels by itself was not sufficient to induce STAT1 gene expression in these cells, but required activation by IFN-␥, as TNF-␣ did not activate the formation of the IGI DNA-binding complex.
The Novel IRF-1-containing Complex Bound to the IGI Element Has a Reduced Mobility Compared with IRF-1-IRF-1 has previously been shown to form a range of different associations, e.g. forming complexes with itself and other members of the IRF family (33)(34)(35), STAT1 (36), and CBP (34,35,37). Given this ability to associate with other proteins and the size of the IGI complex, a study was carried out to further define its components. The mobility in gel shifts of the IGI complex was compared with that produced by preparations of in vitro translated murine IRF-1 protein added to the IGI DNA element as a probe (Fig. 7A). A single dominant band was visible in gel shifts with the in vitro translated IRF-1 protein (indicated as IRF-1 in Fig. 7A), and a band of mobility equivalent to that produced by the in vitro translated recombinant IRF-1 protein was also obtained using the IFN-␥-treated MCF-7 cell samples. Another band of much slower mobility (indicated as IGI complex in Fig.  7A) was also detected in the IFN-␥-treated MCF-7 cell samples.
Agarose beads coupled to the IGI DNA element were used to pull-down specific IGI-binding complexes from extracts of MCF-7 cells treated with IFN-␥. The proteins in the complex were resolved by SDS-PAGE and identified by Western blotting (Fig. 7B). IRF-1 was detected in the IGI-binding complex. The complex also contained CBP, but not STAT1. CBP was detected in the IGI-binding complex only after IFN-␥ treatment.
IFN-resistant Human Melanoma Cell Line MM96 Transfected to Constitutively Express IRF-1 Exhibits Increased Expression of STAT1 and Greater Sensitivity in Biological Assays to IFN-␤-We had previously shown that the human melanoma cell line MM96 is resistant to the effects of IFN-␤ in FIG. 5. A, the GAS site within the IGI element exhibits low level binding to the STAT1 dimer. MCF-7 cells untreated or treated with IFN-␥ for either 30 or 60 min were used to prepare extracts that were combined with radiolabeled probe in the EMSA reaction. The probe used (GAA ACT TTC TGC GAA AAG and its complement) contained the GAS site in a truncated IGI element. Binding of recombinant STAT1 (rSTAT1) was also examined by adding 500 ng of protein o the EMSA reaction. B, mutations in the IRF-E sequences in the IGI element reduce gel shift complex formation. Both the 5Ј-and 3Ј-IRF-Es within the IGI element are important for formation of the novel IGIbinding complex. Extracts for EMSA were prepared from MCF-7 cells that were treated with IFN-␥ for 30 min. Oligonucleotides used as probes in EMSA reactions included the wild-type IGI element (hSTAT1900promIRF-E/GAS/IRF-E) and the IGI element containing mutations within the 5Ј-IRF-E (hSTAT1mut5ЈIRF-E) and the 3Ј-IRF-E (hSTAT1mut3ЈIRF-E). C, mutations in the IGI GAS sequence do not affect complex formation, unlike mutations in the IRF-E sequences, which can eliminate it. SK-MEL-28 cells were treated with IFN-␥ or IFN-␣ for 30 min, and extracts were added to EMSA reactions with radiolabeled probes as follows: the wild-type IGI sequence; the IGI sequence with a mutant GAS element (TTCTGCGAA 3 TGCTAGC-GAA, using the p91-ISRE⌬gas1/2 oligonucleotides); or the IGI sequence with mutant GAAANN regions, IGI mutant 1 (p91-ISREmut1/2) with all three GAAANN sequences mutated and IGI mutant 2 (p91-ISRE-mut3/4) with the two 3Ј-GAAANN sequences mutated. Anti-IRF-1 antibody (Ab) supershifted the IGI complex. biological assays (7,8) as a result of a deficiency in STAT1 protein expression and reduced ISGF3 activity. In addition, cellular resistance to IFN is partly restored by transfecting MM96 cells to constitutively express higher levels of STAT1 protein (7). Analysis of the wild-type MM96 cell line and the STAT1-expressing MM96 subline (7) in the present study by Western blotting revealed that they contained very low levels of IRF-1 protein (Fig 8A). Therefore, the wild-type MM96 cell line was transfected to constitutively express greater levels of IRF-1 protein. Cell extracts from transfected clones were examined for IRF-1 protein expression levels, and a highly expressing clone (c55) was analyzed for cellular levels of STAT1, STAT2, and p48 protein (Fig. 8A). From the results, it can be seen that the MM96 pEF-IRF-1c55 line expressed significantly increased levels of IRF-1 protein, and this was associated with an increased expression of STAT1 to levels similar to those observed in the STAT1-transfected MM96 pEF-STAT1c7 line. In addition, the levels of STAT2 and p48/IRF-9 protein expression were increased to a similar extent in both MM96 sublines transfected to express STAT1 or IRF-1 protein. The response of the MM96 pEF-IRF-1 line to IFN-␤ was examined by antiviral assay (Fig. 8B) and compared with that of the wild-type cell line. The MM96 pEF-IRF-1 line exhibited increased sensitivity to IFN-␤ (CPE 50 , 50% inhibition of cytopathic effect of 5 IU/ml compared with 280 IU/ml for the wild-type cell line). Together, the results from Fig. 8 are consistent with a role for IRF-1 acting in vivo as a key regulator of the human STAT1 gene. DISCUSSION The STAT1 gene was mapped from the CEPH human YAC library, and clones containing the 5Ј-genomic region were isolated by a PCR-based screening approach. Two positive YAC clones (809e2 and 805c1) containing the STAT1 gene were identified (Fig. 1). Using bubble PCR, the 5Ј-genomic region of the STAT1 gene has been characterized and extends previous observations of a closely related organization in the human STAT1 and STAT2 5Ј-genomic regions (38). Thus, both genes have an intron adjoining near the exon containing the ATG initiation codon and an additional 5Ј-exon containing 50 bases (for STAT2). However, the STAT1 gene is distinct in that it contains two additional 5Ј-exons upstream of the exon containing the start codon ( Fig. 2A). We mapped and sequenced the 5Ј-region including exon 1/intron 1/exon 2/intron 2/exon 3. The second intron of the human STAT1 gene is ϳ2.8 kb in size. In the STAT2 gene, the first intron is ϳ3 kb in size. Thus, STAT1 and STAT2 genes show close similarities in genomic organization up to this point. However, the human STAT1 gene contains an additional 5Ј-noncoding exon. FIG. 7. A, the novel IRF-1-containing complex bound to the IGI enhancer exhibits reduced mobility compared with the complex formed by addition of recombinant IRF-1 alone. EMSA was performed using the IGI probe with either extracts from MCF-7 cells treated with IFN-␥ for 30 min or with in vitro translated (IVT) murine IRF-1 protein. Complexes were resolved on 6% polyacrylamide gels and detected by autoradiography. *, nonspecific DNA-binding protein complex. B, CBP binds to the IGI element in an IFN-␥ inducible manner. Neutravidin-agarose saturated with biotinylated IGI oligonucleotides were used to pull-down DNA-bound proteins from extracts of MCF-7 cells untreated or treated with IFN-␥ for 30 or 60 min. Beads were then washed extensively to remove nonspecifically bound proteins before resuspension in SDS-PAGE loading buffer. IGI-binding proteins were resolved by 10% SDS-PAGE and identified by Western blotting using specific antibodies as indicated. I.P., immunoprecipitated.
FIG. 8. A, increased human STAT1 expression in MM96 human melanoma cells transfected to express high levels of IRF-1. The wild-type MM96 human melanoma cell line (7,8) expressing low levels of STAT1 and IRF-1 was transfected with the pEF-IRF-1 vector, and clones were selected for growth in G418. The highly IRF-1-expressing clone (c55) was further analyzed by immunoblotting and compared with wild-type and pEF-hSTAT1-transfected MM96 cell lines. Equal amounts of total cellular protein from lysates of 1 ϫ 10 5 cells prepared in SDS-PAGE sample buffer were examined for the presence of ISGF3 components. The blots were stripped and reprobed with anti-GAPDH antibodies to confirm that sample loadings in each lane were equivalent. B, antiviral assay of IFN-␤ in MM96 and MM96 pEF-IRF-1c55 cell lines. Shown are the dose-response curves for protection against Semliki Forest virus challenge afforded by treating wild-type (⅜) compared with pEF-IRF-1-transfected MM96 cells (f) as indicated over a range of IFN-␤ concentrations.
We isolated a 5Ј-genomic region of 1223 bp from the STAT1 gene, STAT1900prom (Figs. 3 and 4), amplified using EcoRIdigested 805c1 YAC DNA as a template for bubble PCR with the STAT1upstream primer. The sequence obtained from this DNA fragment overlapped with the reported 5Ј-end of the STAT1 cDNA (GenBank TM /EBI accession number M97935). The discrepancy between our sequence and 12 nucleotides at the 5Ј-end of the published cDNA sequence is most probably the result of artifacts generated while preparing the cDNA library reported previously (27). This would also explain why the STAT1upRev primer, mismatched at the four 3Ј-bases, failed to produce a bubble PCR product, whereas use of the STAT1upstream primer gave rise to the correct product. Further support that accession number M97935 has an incorrect 5Ј-end comes from the finding that the next 29 nucleotides in this cDNA sequence (27) are inverted in the reverse direction compared with our DNA sequence (Fig. 2). Analysis of the transcriptional start site using RLM-RACE identified the 5Јend indicated as ϩ1 in Fig. 2B. However, it is unlikely that the human STAT1 gene has a unique start site. An examination of the 5Ј-dbEST Database of Expressed Sequence Tags in the GenBank TM /EBI Data Bank revealed the presence of 21 human STAT1 5Ј-dbESTs, and with the inclusion of our 5Ј-end, 18 of the 22 start at different sites, mostly within ϩ20 bp of the site we have identified. Thus, it is highly probable that the human STAT1 gene uses multiple transcriptional start sites.
Analysis of the human STAT1 intron 1 sequence revealed putative regulatory elements, and this was confirmed when transient transfection assays and EMSA studies were undertaken. All three cell lines (MCF-7, U293T, and SK-MEL-28) transfected with pGL3basic constructs containing the 5Ј-half of STAT1900prom including intron 1 (5Ј-luc) displayed IFN-induced increases in expression of the luciferase reporter gene (Fig. 3), indicating the presence of an IFN-regulated enhancer. Intriguingly, this region also resulted in significant constitutive luciferase activity in the absence of IFN treatment, ϳ10fold above the basal level produced in cells transfected with the pGL3basic control vector alone. The luciferase activity resulting from the 5Ј-luc construct could be further elevated 2-3-fold by treating transfected cells with either type of IFN. It is likely that this constitutive gene transcription results from Sp1 binding to several of the numerous GC box elements located in the 5Ј-end, as Sp1 sites are commonly found in the regulatory regions of ISGs.
By contrast, pGL3 containing the 3Ј-half of STAT1900prom (3Ј-luc) resulted in a 10-fold reduction in the levels of luciferase activity, below those produced by the pGL3basic vector in two of the three cell lines (U293T and SK-MEL-28). These data indicate the presence of a strong repressor or negative regulatory sequence within this 3Ј-half of the STAT1900prom region active in some cell types. Further investigation is required to define the nature of this negative regulatory mechanism and to explain why no suppression was detected in the MCF-7 cell line with 3Ј-luc. Placing the two halves of the STAT1 intronic enhancer region together in the full-length construct restored control to the 5Ј-half because the full-length construct exhibited tight regulation with similar levels of luciferase activity compared with the pGL3basic vector in the absence of IFNs as well as a 5-10-fold induction after treatment with IFN-␣ or IFN-␥ (Fig. 3). These results suggest that the 3Ј-repressor site is important for maintaining tight IFN-inducible regulation from the 5Ј-half of the STAT1 intronic enhancer region. The poor IFN response obtained with either the IGI-luc construct (containing the IGI element alone) or the 5Ј⌬IGI-luc construct (in which only the IGI element was deleted) suggests that the presence of additional sequences in intron 1 act together with the IGI element and that both are required before IFN-activated transcription can occur. The identity of these other sites remains to be determined.
Nevertheless, the IGI DNA element containing IRF-E GAAANN repeats separated by an intervening GAS site in intron 1 was found in gel mobility shift assays to bind a novel complex induced by IFN-␣ and IFN-␥ in all three cell lines examined. In addition, this IGI-binding complex migrated relative to the adjacent reference lane containing the ISG15 ISRE with a slower mobility than that of the prototypic IRF-1 complex (39), but not as slow as the ISGF3-bound complex (data not shown). The IGI-binding complex was also slower than the STAT1 homodimeric complex bound to the human IRF-1 GAS element (Fig. 4A). The supershift of the IGI complex observed with anti-human IRF-1 antibodies, but not with any of the other anti-IRF and anti-STAT antibodies tested (Fig. 5, B and  C), indicates that the IGI complex is unusual in that it contains IRF-1, but does not appear to contain any of the other common IFN-activated transcription factors. No ISGF3 or other complexes of molecules containing p48 or STAT1, STAT2, or STAT3 were visible or could be detected by EMSA or supershift experiments. Although the GAS site within the IGI element was capable of low level binding to the STAT1 dimer in in vitro EMSA reactions, this complex was not detected using extracts from IFN-␥-treated cells (Fig. 5A), and no STAT1 homodimer binding to the 3Ј-GAS site was detected (Fig. 4A). The inability to detect STAT1 or ISGF3 binding to the ISRE/GAS-like STAT1 sequence was surprising given that two GAS half-sites exist in the sequence TTCTGCGAA (15). In addition, mutations introduced to eliminate the IGI GAS site did not affect the DNA binding of the IGI complex induced by IFN-␥ and IFN-␣ (Fig. 5C).
The consensus sequence for IRF-1 binding to DNA has been characterized as a repeat of the GAAANN hexameric oligonucleotide, common to the promoters of the type I IFN genes, the IFN-␥ gene, and other ISGs, usually present as a direct repeat (reviewed in Ref. 41). Crystallography studies revealed that IRF-1 binds as a protein dimer to the GAAA repeat (41)(42)(43), and IRF-1 has been shown to form dimers in vivo (44). In support of this proposal, purified IRF-1 at high concentrations was shown by EMSA to form a novel complex called ISGF2 that bound to the ISRE and that migrated with a slower mobility than the prototypic IRF-1 band (39). The ISGF2 band migrates with similar mobility compared with the larger IRF-1-containing complex (IGI) identified here. Whether these two bands are related remains to be clarified.
With few exceptions, changes to the GAAANN (NNTTTC) repeat motif have been shown to result in a loss of IRF-1 binding and IFN-induced activation (45)(46)(47)(48)(49)(50)(51)(52)(53). Our results were consistent with these observations. In the context of the fulllength STAT1900prom region of DNA, mutation of all three GAAANN repeats within the IGI element resulted in the complete loss of IFN responsiveness (Fig. 3B). All the GAAANN site mutations assayed by EMSA for their effect on IGI complex formation resulted in reduction or loss of IGI binding. It was of interest that mutation of either the 5Ј-or 3Ј-IRF-E resulted in only a partial reduction in IGI binding, whereas mutations of both IRF-Es caused complete loss of binding (Fig. 5, B and C). Thus, the GAAANN repeat sequences are essential for IGI DNA binding of IRF-1 protein complexes. Many studies have demonstrated the existence of functional "imperfect" ISREs, and some also adjoin GAS elements in a arrangement similar to that of the IGI element. In relation to the present study, one of the most interesting is the Sp100 gene promoter, which is also induced by both IFN types and contains two GAS sites and an unusual ISRE (GGAAAAGAGAAGAGAAAGT) that binds preferentially to IRF-1 and only weakly to ISGF3 (54). Thus, the Sp100 promoter resembles the human STAT1 intronic enhancer IGI element with its related sequence, GAAAAGAA-GAAAAC, which we have shown binds IRF-1 preferentially. These two promoters must be distinguished from the more common ISGs such as the guanylate-binding protein gene, in which an ISRE and an overlapping GAS element are both necessary for cooperative transcriptional induction by type I IFN (55).
The observed increase in IRF-1 protein levels in U293T and SK-MEL-28 cells (but not MCF-7 cells) (cf. Figs. 4 and 5 with Fig. 6) was interesting given that IFN-␣ treatment induced increased STAT1 mRNA expression in all three cell lines. It follows that an IGI-independent signaling pathway exists in MCF-7 cells by which IFN-␣ can up-regulate STAT1 expression. Although TNF-␣ treatment of MCF-7 cells raised their IRF-1 protein levels, no associated increase in STAT1 mRNA expression or IGI complex formation occurred. Thus, the TNF-␣ signaling pathway is not sufficient to activate the IGI complex even though it does increase IRF-1 expression levels. By contrast, IFN-␥ was the strongest inducer of IRF-1 protein expression as well as in promoting the formation of the IGI complex in MCF-7 cells. IFN-␥ treatment of these cells was also associated with much greater increases in STAT1 transcription levels. It can be concluded from these results and from EMSA studies with the IGI element that IRF-1-containing IGI complex formation and activation play a key role in the increased human STAT1 mRNA expression in cells treated with IFN-␥.
The histone acetylase proteins CBP/p300 have recently been shown to be important in binding to IRFs and in recruiting them to chromatin to activate transcription (34,35). Our analysis using IGI-conjugated beads in pull-down experiments with cell extracts revealed that the IGI complex also contained CBP after treating cells with IFN-␥ (Fig. 7B). This observation is consistent with the proposal that IFN-␥ signaling activates IRF-1 to bind the IGI element within the STAT1 intronic enhancer in complex with CBP, thereby turning on RNA transcription from the STAT1 gene. Additional evidence that IRF-1 regulates the STAT1 gene in vivo was obtained with the transfected MM96 melanoma cell line expressing IRF-1 protein, as these cells showed elevated expression of STAT1 and other ISGF3 proteins as well as an accompanying enhanced responsiveness to IFN-␤. Thus, our evidence in toto supports a key role for IRF-1 in regulating human STAT1 expression (40). Given that STAT1 regulates the IRF-1 gene promoter (30), we propose that these two genes and their encoded proteins form an intracellular amplifier circuit or feedback loop that acts to regulate cellular responsiveness to the IFNs. The elucidation of reciprocating regulation involving the IRF-1 and STAT1 genes and encoded proteins, two pivotal transcription factors in IFN signaling, provides novel insight into the mechanisms regulating cellular responsiveness to the IFNs. This understanding will also help in further identifying the basis of the changes that take place in cancer cells as they become resistant to IFNs and whereby abnormalities in the expression of key IFN signaling molecules have been elucidated (7)(8)(9)(10)(11).