Stat1 Depends on Transcriptional Synergy with Sp1*

STAT (signal transducer and activator of transcrip- tion) proteins combine with cytokine receptors and re-ceptor-associated kinases in distinct protein/protein in- teractions that are critical for STAT-dependent signal transduction events, but the nature of any subsequent STAT interactions with DNA-binding proteins in the nucleus is less certain. Based on assays of DNA/protein binding and activity of transfected reporter plasmids, we determined that occupation of contiguous DNA-bind- ing sites for Stat1 (the first member of the STAT family) and the transcriptional activator Sp1 are both required for full activation of the intercellular adhesion mole-cule-1 gene by interferon- (cid:103) . Thus, Stat1 binding to DNA cannot by itself be equated with biologic actions of Stat1. In co-immunoprecipitation experiments, we also obtained evidence of direct and selective Stat1/Sp1 in- teraction (in primary culture cells without overexpres-sion), further indicating that Stat1/Sp1 synergy confers an element of specificity in the pathway leading to cy-tokine-activated transcription and cytokine-dependent immunity and inflammation. STAT 1 proteins act as critical intermediates in cytokine-de-pendent gene activation based on their dual capacities for signal transduction (at the cell surface) and activation of transcription (in the nucleus) (1). Signal transduction depends on programmed assembly of cytokine receptors, receptor-associ-ated JAK kinases, and in some cases serine kinases, that recruit and activate specific STAT proteins (2–4). Phosphoryl-ated/activated STATs then dimerize, translocate to the nucleus, and direct transcription of

STAT 1 proteins act as critical intermediates in cytokine-dependent gene activation based on their dual capacities for signal transduction (at the cell surface) and activation of transcription (in the nucleus) (1). Signal transduction depends on programmed assembly of cytokine receptors, receptor-associated JAK kinases, and in some cases serine kinases, that recruit and activate specific STAT proteins (2)(3)(4). Phosphorylated/activated STATs then dimerize, translocate to the nucleus, and direct transcription of specific target genes. For example, the first member of the STAT family (designated Stat1␣) undergoes tyrosine 701 and serine 727 phosphorylation in response to . This activation step is triggered by IFN-␥-dependent oligomerization of the IFN-␥ receptor and consequent cross-phosphorylation of receptor-associated Jak1 and Jak2 kinases and the receptor ␣-chain (6). Receptor phosphorylation enables ␣-chain recruitment of Stat1 via its SH2 domain. Stat1 then undergoes phosphorylation and release from the receptor as a homodimer that can translocate to the nucleus and bind to a specific DNA element (7,8). Thus, distinct protein/protein interactions are critical for Stat1-dependent signal transduction events at the IFN-␥ receptor, but the nature of Stat1 interactions with other proteins (especially other transcription factors) in the nucleus is less certain. In the present report, we take advantage of a primary cell culture model with selective IFN-␥ responsiveness of the intercellular adhesion molecule-1 (ICAM-1) gene (9,10) in order to study the basis for Stat1-dependent transcription. The results offer the first evidence that Stat1-mediated transactivation depends on synergistic interaction with another transcriptional activator (Sp1).
Gel Mobility-shift and DNA-Protein Cross-linking-Our methods for isolation and culture of hTBEC monolayers, preparation of hTBEC cellular and nuclear proteins, and gel mobility-shift assays have been described previously (9,10,12). Gel mobility-shift assays were modified to include UV-induced DNA-protein cross-linking and analysis of crosslinked complexes by SDS-PAGE. To enhance UV-induced cross-linking, double-stranded BrdUrd-containing oligonucleotides were generated by using the first strand as a template for second strand synthesis with the Klenow fragment of Escherichia coli DNA polymerase I (Promega), an octamer primer, a dNTP mixture containing BrdUrd (Sigma) in place of thymidine, and [␣-32 P]dCTP. Nuclear protein/oligonucleotide mixtures initially underwent standard gel mobility-shift assay, and the shifted bands containing 32 P-labeled complexes were localized by overnight autoradiography at 4°C, excised, and subjected to UV-irradiation ( ϭ 254 nm). The gel slices were then equilibrated with sample buffer, boiled, and subjected to SDS-PAGE with 10% polyacrylamide. The resolved DNA-protein complexes were electrophoretically transferred to PVDF membranes (Immobilon-P, Millipore Corp., Bedford, MA) and subjected to autoradiography. To determine if the DNA-protein complexes contained Stat1␣ and/or Stat1␤, PVDF membranes containing resolved complexes were treated with 15% hydrogen peroxide for 10 min, blocked with 5% nonfat milk in Tris-buffered 0.9% NaCl with 0.1% Tween 20, and then incubated with p91 and/or p91T antisera. Primary Ab binding was then detected using anti-rabbit horseradish peroxidase conjugate (Boehringer Mannheim) and an enhanced chemiluminescence method (Amersham International).
Reporter Plasmids-The luciferase-reporter plasmid with ICAM-1 gene sequence (Ϫ130 to Ϫ35) driving the Photinus pyralis luciferase gene (pBH-130 -35ICAM-1-luc) was described previously (10). ICAM-1 (Ϫ130 to Ϫ35) with an inverted repeat mutation was generated by PCR using BamHI-digested pBH-1335ICAM-1/MUT1-luc (which contains the sequence GACCTCTTACA in place of wild-type inverted repeat TTTCCGGGAAA) as DNA template (10), a downstream primer containing a HindIII restriction site and ICAM-1 sequence Ϫ64 to Ϫ35, and an * This research was supported by Grants HL/AI-51071, HL-40078, HL-07317, and HL-02920 from the National Institutes of Health and a grant from the American Lung Association. 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.
upstream primer with a KpnI site and ICAM-1 sequence Ϫ130 to Ϫ102 with the same inverted repeat mutation. ICAM-1 (Ϫ130 to Ϫ35) with a GC box mutation was generated by PCR using BamHI-digested pBH-176ICAM-1-luc as template (13), the same downstream primer as for the inverted repeat mutation, and an upstream primer containing a KpnI site and ICAM-1 sequence Ϫ130 to Ϫ76 with the sequence CCGAAC replacing the putative Sp1 sequence CCGCCC at Ϫ99 to Ϫ94. PCR products were digested with KpnI and HindIII, purified by electrophoresis, and ligated into KpnI/HindIII-digested pBH-luc (13). The luciferase-reporter plasmid with ICAM-1 gene sequence (Ϫ130 to Ϫ94) and the minimal promoter sequence from the Herpes simplex thymidine kinase gene (pBH-130 -94ICAM-1-TK-luc) was described previously (10). Corresponding plasmids with mutations of the ICAM-1 gene inverted repeat or GC box were generated by synthesis of double-stranded oligonucleotides containing the wild-type or mutant ICAM-1 sequence, an upstream NotI, and a downstream XhoI site followed by ligation into NotI/XhoI-digested pBH-TK-luc. The ICAM-1-Gal4-luciferase-reporter plasmid (pBH-130 -35ICAM-1/Gal4-luc) was generated by ligating a double-stranded oligonucleotide with two Gal4 sites (14,15) interposed between two XhoI sites into an XhoI-digested plasmid (pBH-130 -35ICAM-1/XhoI-luc) with an XhoI site in place of the GC box. Insert orientation and sequence integrity of all plasmid constructs was verified by DNA sequencing.
Cell Transfection and Reporter Gene Assay-Cell transfection and reporter gene assays (for luciferase and CAT activities) were performed as described previously (10). For cotransfection experiments, the levels of reporter and expression plasmids were titered to provide similar levels of basal promoter activity among experiments (7.5 g of reporter plasmid and 0.1-2.5 g of expression plasmid were used per 60-mm culture plate).
Co-immunoprecipitation-Whole cell protein extracts were prepared by lysing hTBECs in 50 mM Tris, pH 8, 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 750 M dithiothreitol, 1 mM sodium orthovanadate, 10 mM sodium fluoride, 2 mM sodium pyrophosphate, 10 g/ml leupeptin, and 10 g/ml aprotinin. For co-immunoprecipitations, 200 g of cell protein was incubated for 1 h at 4°C with 5 g of primary Ab in 500 l of lysis buffer. Control Ab was an immunoaffinity-purified antiserum against an irrelevant peptide. Immune complexes were precipitated with Protein A-Sepharose and then immunblotted and detected as described for DNA-protein cross-linking experiments.

RESULTS AND DISCUSSION
To investigate the basis for STAT-dependent transcription, we used a primary culture epithelial cell (hTBEC) model that exhibits IFN-␥-dependent ICAM-1 expression under the control of an IRE at nucleotides Ϫ130 to Ϫ94 of the ICAM-1 gene (10). This IRE contains an inverted repeat (at Ϫ116 to Ϫ106) that is critical for forming an IRE-binding complex (IRE-BC) and for IFN-␥ responsiveness of the ICAM-1 gene, and this inverted repeat is similar to a motif found in several other IFN-␥-responsive genes (7,20,21). In each case, the inverted repeat motif is sufficient for binding phosphorylated Stat1 (11,(22)(23)(24)(25), and this capacity is also exhibited by the inverted repeat in the ICAM-1 gene in transformed cell lines (26 -28). DNA-protein cross-linking experiments using the ICAM-1 gene inverted repeat and nuclear proteins from IFN-␥-stimulated hTBECs indicate that Stat1 directly contacts the DNA at this site (Fig. 1).
Although a series of IFN-␥-responsive genes share a capacity to bind Stat1, it is not clear that Stat1 binding alone is sufficient for full responsiveness to IFN-␥. For example, deleting DNA elements surrounding the inverted repeat leads to a marked decrease in the maximal level of responsiveness to IFN-␥ that is often linked to a concomitant decrease in (cyto- FIG. 1. Activated Stat1 binds directly to the ICAM-1 gene inverted repeat. Nuclear proteins from IFN-␥-stimulated hTBECs (100 units of IFN-␥/ml for 1 h) were subjected directly to immunoblotting with anti-Stat1 Ab (lane 1) or were mixed with 32 P-labeled, BrdUrdcontaining ICAM-1 gene sequence (Ϫ120 to Ϫ98). The mixture underwent nondenaturing 5% PAGE, and the 32 P-labeled band was excised, treated with or without UV irradiation, and then subjected to 10% SDS-PAGE and electrophoretic transfer to PVDF membranes for anti-Stat1 Ab immunoblotting (lanes 2 and 3) and 32 P autoradiography (lanes 4 and 5). The slower migrating Stat1 bands represent Stat1␣ and Stat1␤ covalently bound to the ICAM-1 gene inverted repeat (designated Stat1␣/IR and Stat1␤/IR), and the faster migrating bands represent unbound Stat1␣ and Stat1␤.

FIG. 2. Sp1 binds 3-adjacent to Stat1 in the ICAM-1 gene IRE.
A, design for wild-type (WT) and mutant (MU) ICAM-1 gene sequences that were tested for their capacity to bind nuclear proteins from unstimulated and IFN-␥-stimulated hTBECs in gel mobility-shift assays. B, wild-type ICAM-1 sequences prepared as 32 P-labeled, doublestranded oligonucleotides of varying lengths were used to define binding sites mediating formation of two IRE-binding complexes (IRE-BC 1 and IRE-BC 2 ). Note that the GC box alone is necessary but not sufficient for Sp1-binding (lanes 1 and 2 versus lanes 3-6) so that some less specific adjoining sequence is also required. C, ICAM-1 sequence (Ϫ130 to Ϫ81) without or with inverted repeat or GC box mutations were used to define binding sites for Stat1 and Sp1. D, nuclear protein extracts from unstimulated hTBEC monolayers were treated with no antiserum (lane 1) or with Abs against Sp1 (lane 2), AP-2 (lane 3), Stat1 (lane 4), or nonimmune IgG (lane 5) and then mixed with 32 P-labeled oligonucleotides containing ICAM-1 gene sequence (Ϫ130 to Ϫ81). Arrows indicate the positions of binding and supershifted (*) complexes. E, nuclear protein extracts were prepared from IFN-␥-stimulated hTBEC monolayers and assayed for DNA binding as in D.

Stat1/Sp1 Transcriptional Synergy 30265
kine-independent) basal promoter activity (10,23,25,27). Basal activity and IFN-␥ responsiveness are restored when the Stat1-binding site is placed in front of a heterologous promoter (10), but these systems may include other DNA elements that could synergize with the Stat1 site (as noted below). We were therefore interested in further analyzing the types of DNA/ protein and protein/protein interactions in the ICAM-1 gene promoter region that serve to fully activate IFN-␥-driven transcription.
A distinct feature of the ICAM-1 gene IRE is the presence of a putative binding site (a GC box at Ϫ99 to Ϫ94) for the Sp1 transcription factor (29) at the 3Ј-end of the element ( Fig. 2A). Accordingly, we performed gel mobility-shift assays using the ICAM-1 gene inverted repeat and the adjacent GC box. When the GC box and adjoining sequence were included in the oligonucleotide probe, a second more slowly migrating complex (designated IRE-BC 2 ) was observed (Fig. 2B). This behavior fits with observations that: (i) in general, the GC box alone is necessary but not sufficient for Sp1-binding so that some less specific adjoining sequence is also required (30); and (ii) for the ICAM-1 gene, the footprint for Sp1 extends to nt Ϫ81 by DNase I protection mapping. 2 Formation of IRE-BC 2 was decreased by mutation of the GC box but not by mutation of the inverted repeat, whereas formation of the original, faster migrating complex (designated IRE-BC 1 ) was decreased by inverted repeat but not by GC mutation (Fig. 2C). In addition, IRE-BC 2 was present in unstimulated and IFN-␥-stimulated cells and was supershifted with anti-Sp1 Ab, whereas IRE-BC 1 was present only in IFN-␥-stimulated cells and was reactive only with anti-Stat1 Abs (Fig. 2, C and D). Thus, two distinct sites in the ICAM-1 gene IRE (an inverted repeat at Ϫ116 to Ϫ106 and a GC box at Ϫ99 to Ϫ94) are responsible for formation of two DNA-protein complexes: IRE-BC 1 , which contains Stat1; and IRE-BC 2 , which contains Sp1.
The functional significance of Sp1-and Stat1-binding sites A, hTBECs were transfected with a luciferase-reporter plasmid containing ICAM-1 gene sequence (Ϫ130 to Ϫ35) or the same sequence with mutations in the inverted repeat (Stat1-binding site) or GC box (Sp1-binding site). After transfection, cells were incubated without (open bars) or with (dark bars) IFN-␥ and then assayed for luciferase activity. B, the same transfection protocol was followed using constructs with ICAM-1 gene IRE sequence (Ϫ130 to Ϫ94) with or without mutation of the inverted repeat or GC box driving a heterologous minimal promoter (TK, thymidine kinase). C, cells were transfected with reporter plasmids containing ICAM-1 gene sequence (Ϫ130 to Ϫ35) or the same sequence with the GC box replaced by Gal4-binding sites. Cells were cotransfected with expression plasmids for Gal4 DNA-binding domain alone (Gal4) or with this domain fused to the transactivating domain(s) for Sp1 (Gal4/Sp1) or p65 (Gal4/p65). Results for Gal4/Fos and Gal4/E1a chimeras were similar to those for Gal4/p65, and expression level of each chimeric protein was verified by immunoblot against anti-Gal4 Ab (data not shown). In A-C, each value is the average of duplicate samples and is representative of three experiments; similar values were obtained in experiments with pRSV-CAT cotransfection to control for transfection efficiency (data not shown).

Stat1/Sp1 Transcriptional Synergy 30266
for IFN-␥-responsiveness of the ICAM-1 gene was determined using a series of ICAM-1/luciferase-reporter gene constructs in transient transfection assays of unstimulated versus IFN-␥stimulated epithelial cells. Reporter constructs contained either wild-type ICAM-1 gene 5Ј-flanking sequence or the same sequence with mutation of the Stat1-binding inverted repeat (Ϫ116 to Ϫ106) or the Sp1-binding GC box (Ϫ99 to Ϫ94). Mutations that blocked protein binding to either site also blocked IFN-␥ responsiveness in transfection experiments using constructs containing ICAM-1 gene sequence from Ϫ130 to Ϫ35 (i.e. using the ICAM-1 gene endogenous promoter, Fig.  3A) or ICAM-1 sequence from Ϫ130 to Ϫ94 fused to a minimal promoter construct containing the Herpes simplex thymidine kinase gene promoter (i.e. using a heterologous promoter, Fig.  3B). Thus, a functional Sp1 site (located close to the transcription initiation site) was required for full IFN-␥ responsiveness.
To determine whether the effect of Sp1 was specific or could be conferred by any transcriptional activator, Sp1 and a series of activators were brought to the promoter as fusion proteins containing a Gal4 DNA-binding domain. Direct evidence that Sp1 is needed for full IFN-␥ responsiveness was obtained when co-transfection of a Gal4/Sp1 fusion protein restored full responsiveness to a relatively inactive construct containing ICAM-1 5Ј-flanking sequence with a Gal4-binding site in place of the Sp1-binding site (Fig. 3C). Specificity for Sp1 was indicated when its capacity to restore full responsiveness was not shared with the transactivating domains of several other transcription factors (including p65/RelA, Fos, or E1a), even though each of these factors was capable of increasing basal promoter activity to a similar level (Fig. 3C). Taken together, the results indicate that both Stat1 and Sp1 binding are required for maximal IFN-␥ activation of the ICAM-1 gene. The proximity of the two binding sites and the selective effect of Sp1 also suggested a specific interaction between the two transcription factors.
Accordingly, we next examined the capacity of Sp1 and Stat1 for direct protein/protein interaction. Co-immunoprecipitation experiments using either nuclear or whole cell extracts from unstimulated or IFN-␥-stimulated cells indicated direct Sp1/ Stat1 interaction (Fig. 4, A and B). This interaction was therefore not dependent on Stat1 phosphorylation state. Additional immunoblotting indicated that hTBECs also contain Stat2, -3, -5, and -6; however, only Stat3 appeared to directly interact with Sp1 (data not shown). This finding correlates with a higher degree of homology between Stat1 and Stat3 than among other STAT family members (31)(32)(33)(34)(35)(36). Immunofluorescence microscopy demonstrated that Stat1 was localized to the cytosol in unstimulated cells and to the nucleus in IFN-␥stimulated cells, whereas Sp1 was found in the nucleus under both conditions (data not shown). Predominant Sp1/Stat1 interaction must therefore occur in the nucleus and is limited by nuclear translocation of Stat1.
In summary, we find that maximal IFN-␥-driven activation of the ICAM-1 gene depends on two critical DNA/protein interactions: 1) an inverted repeat motif that mediates direct binding of activated/phosphorylated Stat1, and 2) a GC box motif that recruits constitutively active Sp1. We also present evidence for Stat1/Sp1 interaction that may serve to confer an additional element of specificity for transcriptional activation (Fig. 4, C and D). In this model, Sp1 may help recruit Stat1 to the promoter and/or may serve to better link Stat1 action to the basal transcription complex. Interestingly, IFN-␥-dependent activation of a second immune response gene in airway epithelial cells, the interferon regulatory factor-1 gene, is also Stat1dependent 2 and also contains a GC box with a corresponding DNA footprint downstream of the inverted repeat motif (20,37). Thus, IFN-␥-driven activation of an immune response subset may be linked by a signal transduction pathway that includes transcriptional synergy between Stat1 and Sp1. Singlecytokine activation of this subset in the epithelial barrier provides an efficient molecular basis for mediating mucosal immunity and inflammation.