Interferon gamma-dependent induction of human intercellular adhesion molecule-1 gene expression involves activation of a distinct STAT protein complex.

In response to interferon γ (IFNγ), intercellular adhesion molecule-1 (ICAM-1) is expressed on human keratinocytes, a cell type that is critically involved in cutaneous inflammation. An ICAM-1 5′ regulatory region palindromic response element, pIγRE, has been shown to confer IFNγ-dependent transcription enhancement. By electrophoretic mobility shift assays (EMSA), pIγRE forms a distinct complex with proteins from IFNγ-treated human keratinocytes, termed γ response factor (GRF). Binding of GRF is tyrosine phosphorylation-dependent, and mutations of pIγRE that disrupt the palindromic sequence or alter its spatial relationship abrogate GRF binding. Supershift EMSAs using antibodies to characterized STAT proteins suggest that GRF contains a Stat1α-like protein; however, non-ICAM-1 IFNγ-responsive elements (REs) known to bind Stat1α homodimers fail to compete for GRF binding in EMSA, and pIγRE does not cross-compete with these REs that complex with homodimeric stat1α. The pIγRE·GRF complex also displays a distinctly different electrophoretic mobility compared to that of IFNγREs complexed to homodimeric Stat1α. These findings indicate that a distinct complex containing a Stat1α-like protein mediates IFNγ-induced ICAM-1 gene transcription and identifies a subset of IFNγ-responsive genes that appear to be regulated by this complex.

In response to interferon ␥ (IFN␥), intercellular adhesion molecule-1 (ICAM-1) is expressed on human keratinocytes, a cell type that is critically involved in cutaneous inflammation. An ICAM-1 5 regulatory region palindromic response element, pI␥RE, has been shown to confer IFN␥-dependent transcription enhancement. By electrophoretic mobility shift assays (EMSA), pI␥RE forms a distinct complex with proteins from IFN␥treated human keratinocytes, termed ␥ response factor (GRF). Binding of GRF is tyrosine phosphorylation-dependent, and mutations of pI␥RE that disrupt the palindromic sequence or alter its spatial relationship abrogate GRF binding. Supershift EMSAs using antibodies to characterized STAT proteins suggest that GRF contains a Stat1␣-like protein; however, non-ICAM-1 IFN␥-responsive elements (REs) known to bind Stat1␣ homodimers fail to compete for GRF binding in EMSA, and pI␥RE does not cross-compete with these REs that complex with homodimeric stat1␣. The pI␥RE⅐GRF complex also displays a distinctly different electrophoretic mobility compared to that of IFN␥REs complexed to homodimeric Stat1␣. These findings indicate that a distinct complex containing a Stat1␣-like protein mediates IFN␥-induced ICAM-1 gene transcription and identifies a subset of IFN␥-responsive genes that appear to be regulated by this complex.
The initiation and evolution of localized inflammation is a consequence of homing and extravasation of leukocytes at sites of tissue injury or threat (1). Intercellular adhesion molecule-1 (ICAM-1) 1 , a cell surface glycoprotein that belongs to the im-munoglobulin gene superfamily, serves as a specific ligand for receptors expressed by leukocytes (2). ICAM-1 plays a pivotal role in the adhesion and transmigration of leukocytes at sites of inflammation, and the up-regulation of ICAM-1 cell surface expression during inflammatory responses is essential in facilitating leukocyte migration (3,4).
IFN␥, a pleiotropic cytokine produced by activated T lymphocytes, plays a critical role in host defenses and inflammation (5). In the skin IFN␥ induces de novo expression of ICAM-1 on human keratinocytes (HK), cells that are critically involved in cutaneous inflammatory processes (6). Upon binding to its receptor, IFN␥ initiates a signal transduction cascade that involves rapid activation of two members of the Janus tyrosine kinase family, JAK-1 and JAK-2 (7), and consequent tyrosine phosphorylation and activation of a latent cytoplasmic protein, originally referred to as p91 and now known as Stat1␣ (8,9). Stat1␣ is the first described member of a family of proteins known as STATs, or signal transducers and activators of transcription. When activated by IFN␥, activated Stat1␣ homodimerizes through Src homology domains (10,11) to form ␥-activated factor (GAF). The Stat1␣ homodimers, after translocation to the nucleus, bind to the ␥-activated site (GAS), first identified in the human guanylate-binding protein (GBP) gene, and initiate gene transcription (8,9,12). Stat1␤ (p84) is an alternatively spliced product of the gene for Stat1 and is a truncated protein that lacks 38 amino acids at the carboxyl end of Stat1. Stat1␤ also homodimerizes and is capable of binding GAS but does not activate transcription (13). It has recently been shown that Stat1␣, which obligatorily requires tyrosine phosphorylation to become active, also requires phosphorylation of a serine residue for maximal activation of gene transcription (14).
In addition to formation of the homodimeric GAF, Stat1 also participates in forming the heteromultimeric transcription complex ISGF3, composed of Stat1␣, stat2, and a non-STAT protein, p48 (15)(16)(17)(18). ISGF3 binds to the IFN-stimulated response element (ISRE), which is an IFN␣/␤and IFN␥-inducible element involved in the regulation of a variety of genes (11,13,17). Recently, multiple cytokines and growth factors have been shown to mediate their transcriptional effects through these and additional STAT proteins, of which there are now six characterized members (19). However, Stat1␣ is frequently activated in response to a wide variety of extracellular signals, including those involved in transcriptional activation of a num-ber of genes involved in immune responses (19).
IFN␥-dependent induction of ICAM-1 gene expression is regulated at the transcriptional level (20,21). The 5Ј flanking region of ICAM-1 gene contains an 11-base pair (bp) element, which we refer to as palindromic IFN␥ response element (RE), or pI␥RE, located upstream of the ICAM-1 transcription initiation site between nucleotides Ϫ76 and Ϫ66. pI␥RE is composed of the sequence 5Ј-TTTCCGGGAAA-3Ј. Several laboratories have demonstrated that pI␥RE is both necessary and sufficient for IFN␥-dependent gene transcription (20,22). 2 The present studies were designed to characterize the molecular events and trans-acting factors involved in the IFN␥induced regulation of ICAM-1 gene transcription. The data presented show that the protein complex activated by IFN␥, which trans-activates ICAM-1 gene expression by binding to pI␥RE, shares both similarities and distinct differences with previously characterized IFN␥-activated STAT complexes. From these data we propose that the protein complex mediating IFN␥-dependent ICAM-1 gene transcription, which we refer to as the ␥ response factor, or GRF, represents a distinct form of IFN␥-induced transcription trans-activator and likely mediates trans-activation of a subset of IFN␥-inducible early response genes.

MATERIALS AND METHODS
Cell Culture-As described previously (22), HK were isolated from neonatal foreskins at the Emory Skin Diseases Research Center. HK were cultured in KGM supplemented with bovine pituitary extract (Clonetics Corp., San Diego, CA). Cultures were maintained at 37°C in humidified 5% CO 2 and passaged at 60 -70% confluence using subculture reagents from Clonetics. Experiments with HK were conducted with cells in passage 3.
Cytoplasmic and Nuclear Extract Preparation-Cytoplasmic and nuclear extracts were prepared as described previously (23) from cells that were either left untreated or treated with 250 units/ml recombinant human IFN␥ (R&D Systems, Minneapolis, MN). Cells were washed twice in ice-cold phosphate-buffered saline (Life Technologies, Inc.), then quickly washed in buffer A (10 mM Hepes, pH 7.4, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM dithiothreitol (DTT), 50 M phenylmethylsulfonyl fluoride (PMSF); all from Sigma). After centrifugation at 2,000 rpm (4°C for 5 min), the cell pellet was resuspended in buffer A containing 0.1% Nonidet P-40 (U. S. Biochemical Corp.) and incubated on ice for 10 min. The nuclei were collected by centrifugation at 4,000 rpm for 2 min at 4°C. The supernatant (cytosolic fraction) was removed and saved, and the pellet was resuspended in buffer B (20 mM Hepes, pH 7.4, 1.5 mM MgCl 2 , 420 mM NaCl, 1 mM DTT, 50 M PMSF, 0.2 mM EDTA, 25% glycerol). The nuclear pellet was incubated on ice for 30 min, followed by centrifugation at 14,000 rpm for 15 min. The protein concentrations of the cytosolic and nuclear fractions were determined using UV absorbance at 280 nm as described (24). Proteins were used immediately in a binding reaction or aliquoted and stored at Ϫ70°C.
Electrophoretic Mobility Shift Assay (EMSA)-The DNA binding reaction was performed for 30 min at room temperature in a volume of 20 l, containing 5 g of nuclear or cytoplasmic protein extract, 2.5 g of bovine serum albumin (Life Technologies, Inc.), 2 g of poly(dI-dC) (Sigma) 5 l of 4 ϫ binding buffer (1 ϫ buffer: 12 mM Hepes, pH 7.8, 4 mM Tris, 60 mM KCl, 1 mM EDTA, 12% v/v glycerol, 1 mM DTT, 1 mM PMSF) with or without 10 -100-fold molar excess of cold competitor DNA. Radiolabeled probe (1 ϫ 10 5 cpm) was added for an additional incubation period of 20 min. In supershift EMSA, cytoplasmic or nuclear extracts were incubated with experimental or isotype control antibody, at supplier's recommended concentrations, prior to the addition of the 32 P-labeled probe. Polyclonal antibodies to Stat1 and monoclonal antibodies to Stat1 through Stat6 and to p48 were obtained from Transduction Laboratories, Lexington, KY. Polyclonal antibodies to both the amino and carboxyl termini of Stat1 were kindly provided by J. E. Darnell, Rockefeller University, New York, NY. DNA binding reactions were separated on 4% native polyacrylamide gels. Gels were subsequently dried and autoradiography performed. The autoradiographs were scanned on a La Cie flat bed scanner (La Cie Ltd., Beaverton, OR) utilizing Adobe Photoshop software (Adobe Systems, Inc., Mountain View, CA). Subsequently the digitized image was labeled in Microsoft Power Point (Microsoft Corp., Redmond, WA) and printed on a high resolution laser printer. Each figure represents a computergenerated image of the autoradiograph, and each is typical of the autoradiograph in the context of relative band and background densities.
To test the requirement of tyrosine phosphorylation for DNA-protein complex formation, nuclear protein extracts were exposed to 10 units of protein-tyrosine phosphatase (PTPase) 1-B from Yersinia pestis (Upstate Biotechnology, Inc., Lake Placid, NY) in the presence or absence of the PTPase inhibitor sodium orthovanadate (1 mM). These reactions were carried out for 30 min at 30°C in a 15-l reaction containing 20 mM Tris-HCl, pH 7.4, 0.5 mM DTT, 0.1 mM EGTA. Treated extracts were then incubated with 32 P-labeled pI␥RE as probe and analyzed by EMSA.
All oligonucleotides used for probes or as cold competitors were synthesized at the Emory University Microchemical Facility. Doublestranded oligonucleotides were prepared by annealing complementary strands as described (25). The pI␥RE oligonucleotide was synthesized to include the IFN␥-responsive site (in bold letters) found in the ICAM-1 gene promoter (5Ј-CGAAGCTTTTCCGGGAAAGGATCCC-3Ј). The underlined sequences in the pI␥RE oligonucleotide represent restriction sites that were used to create overhangs for labeling with 50 Ci of [␣-32 P]dCTP (DuPont NEN) by fill-in reaction using Klenow (Stratagene, La Jolla, CA) as described (25). Alternatively, a short primer, GGGATCCTTTCC, complementary to the 3Ј end of the above sequence was used to label pI␥RE by a primer-extension fill-in reaction using Klenow and 50 Ci of [␣-32 P]dCTP (25). Unincorporated nucleotides were removed by column chromatography over G25 Sephadex columns (Boehringer Mannheim). The [␣-32 P]dCTP-labeled probe was used in analysis of the phosphatase-treated or untreated proteins in EMSA. In all other EMSAs 5Ј end-labeled probes were prepared with 100 Ci of [␥-32 P]ATP (Amersham Corp.) using T4 polynucleotide kinase (New England Biolabs, Beverly, MA) and were gel-purified as described (25).
In experiments comparing complexes and mobility of pI␥RE versus the GAS element, a short primer (GGGATCCGATTTAGA) complementary to the 3Ј end of the GAS sequence of the GBP gene, was used to label GAS by a primer-extension fill-in reaction using Klenow and 50 Ci of [␣-32 P]dCTP. Unincorporated nucleotides were removed by column chromatography over G25 Sephadex columns.
In addition, four oligonucleotides incorporating distinct mutations, insertions, or deletions in the wild type pI␥RE sequence were also used as competitors or probes. These pI␥RE mutants, with mutated nucleotides shown in italics, were: MUT 1, CGAAGCTTCACCGGGAAAG-GATCCC; MUT 2, CGAAGCTTTTCCGGGAGAGGATCCC; MUT 3, CGAAGCTTTTCCATGCATGCATGGAAAGGATCCC; MUT 4, CGAAGCT TTTCC(X)GGAAAGGATCCC. Each mutant incorporates a distinct class of change; MUT 1 has a two-nucleotide mutation in the 5Ј side of the of pI␥RE, MUT 2 has a one-nucleotide mutation in the 3Ј side of pI␥RE, MUT 3 has a 10-bp insertion separating the 5Ј and 3Ј sides of the palindrome, and MUT 4 deletes the central G-C hinge in the palindrome. In addition, two different 30-bp sequences from regions of ICAM-1 located upstream from the pI␥RE were used as irrelevant competitor DNA fragments.

IFN␥ Induces Activation of a Specific pI␥RE-binding Factor (GRF) in a Time-dependent Manner-
We examined by EMSA whether specific factors that can bind to pI␥RE are activated in HK by IFN␥ treatment. As seen in Fig. 1A, stimulation of cells with IFN␥ led to the induction of a distinct DNA-binding protein complex (GRF), resulting in retarded mobility of the labeled pI␥RE probe. However, no pI␥RE binding activity was observed in nuclear extracts isolated from untreated HK. Bind-ing of GRF to labeled pI␥RE was not competed by excess unlabeled, irrelevant DNA as competitor, while excess unlabeled pI␥RE competed for GRF binding, indicating the specificity of the retarded complex. In order to determine the kinetics of induction of GRF, HK were treated with IFN␥ for various periods of time before preparing both cytoplasmic and nuclear extracts for EMSA. As seen in Fig. 1B, the binding activity of cytoplasmic extracts from HK peaked after 30 min of IFN␥ treatment, decreased after 4 h, and declined significantly by 8 h. By 24 h, binding activity was not present. Stimulation of HK with IFN␥ led to the induction of GRF in both cytoplasmic as well as nuclear fractions within seconds (data not shown), consistent with the characteristic rapid activation of latent proteins by this cytokine (13).
pI␥RE Is Necessary for GRF Binding Activity, and Mutated pI␥RE Sequences neither Compete nor Bind to GRF-We next addressed whether targeted mutations, insertions, or deletions within the 11-bp pI␥RE palindromic sequence had any effect upon GRF binding. As seen in Fig. 2A, mutations that disrupted either the 5Ј or the 3Ј end of the pI␥RE palindrome failed to compete with wild type pI␥RE for binding to GRF. Moreover, substitution with a 10-bp spacer (allowing for one full helical turn) between the palindromic halves or deletion of the G-C hinge of the palindrome abrogated the ability to compete with wild type pI␥RE for binding to GRF. When used as radiolabeled probes, these mutated oligonucleotides also failed to display any retarded complexes upon incubation with lysates from IFN␥-treated cells. Because the mutant pI␥RE sequences failed to display any retarded complexes and failed to compete with pI␥RE for GRF complex formation at 10-fold molar excess, we investigated whether these mutants displayed a concentration-dependent competition for GRF at higher molar ratios. Unlabeled mutant oligonucleotides ranging from 10-fold to 1000-fold molar excess were used in a competition EMSA. As seen in Fig. 2C, the mutants failed to display any competition with pI␥RE for GRF complex formation even at a 1000-fold molar excess. Furthermore, unlabeled double-stranded oligonucleotide corresponding to the GAS sequence also failed to display competition for GRF complex formation when used at a 1000-fold molar excess. Finally, while the unlabeled doublestranded irrelevant oligonucleotide also failed to display competition for GRF at a 1000-fold excess, a competition EMSA utilizing unlabeled pI␥RE displayed a detectable diminution of the labeled GRF complex at a unlabeled to labeled ratio of 0.01 to 1 (Fig. 2B), and a significant competition at 0.1 and 1 to 1 molar ratios (Fig. 2, B and C). These data demonstrate a high and specific binding affinity of pI␥RE for GRF binding, a requirement for a specific sequence and spatial arrangement within this element, and a lack of competition by an oligonucleotide containing the classic GAS element previously demonstrated to bind homodimeric Stat1␣.
The changes in the pI␥RE palindromic sequence that abrogated GRF binding and complex formation were consistent with additional transcriptional activation studies using reporter gene constructs. None of the described mutations were capable of driving reporter gene expression, while wild type pI␥RE was sufficient to confer IFN␥ inducibility of heterologous reporter gene constructs (data not shown). These results indicate that GRF requires an intact pI␥RE palindromic sequence to bind to pI␥RE in vitro and pI␥RE is necessary to function in vivo.
GRF Activation Is Dependent upon Tyrosine Phosphorylation-The importance of protein-tyrosine phosphorylation in the activation of trans-acting proteins involved in IFN␥-induced transcription of other genes (19,31) led us to investigate whether pI␥RE binding activity induced upon IFN␥ treatment of HK was dependent upon tyrosine phosphorylation. Purified recombinant PTPase 1-B, isolated from Y. pestis, is a close homologue of the human tyrosine-specific phosphotase PTPase 1 and has been found to specifically catalyze the removal of phosphate from tyrosine residues (32). We assessed the effect of PTPase 1-B on the pI␥RE⅐GRF complex formation. When the pI␥RE probe was incubated with IFN␥-treated HK cell lysates in the presence of PTPase 1-B, formation of the pI␥RE⅐GRF complex was abrogated (Fig. 3). However, addition of the PTPase inhibitor sodium orthovanadate prior to incubation with the enzyme prevented disruption of complex formation, while the inhibitor itself had no effect upon the complex formation. These results indicate that tyrosine phosphorylation is needed for formation of the pI␥RE⅐GRF complex, a property consistent with GRF containing a STAT-like protein (13).
Anti-Stat1␣ Antibodies Interact with the GRF-Since IFN␥ activation of Stat1␣ is known to involve tyrosine phosphorylation of quiescent cytoplasmic proteins, we investigated the possibility of Stat1␣ involvement in GRF using anti-Stat1␣ antibodies in supershift EMSAs. As shown in Fig. 4, addition of polyclonal anti-Stat1␣ antibodies directed against either the amino-terminal or the carboxyl-terminal portions of Stat1␣ to IFN␥-treated cell lysates supershifted the pI␥RE⅐GRF complex, and these supershifted complexes were competed away by excess unlabeled pI␥RE. Irrelevant antibodies had no effect upon complex formation or mobility (data not shown). Since polyclonal antibodies raised against both the amino-and the carboxyl-terminal portions of Stat1␣ supershift the pI␥RE⅐GRF complex, and since Stat1␤ lacks the 38 carboxyl-terminal amino acids that are included in Stat1␣, it is extremely unlikely that the Stat1-like protein identified through these studies is Stat1␤. Monoclonal anti-Stat1␣ antibodies used in subsequent studies, while specific for Stat1␣, were raised against peptide regions common to Stat1␣ and Stat1␤, and thus do not provide evidence to include or exclude Stat1␤ as a possible component of the pI␥RE⅐GRF complex. However, polyclonal antibody supershift data presented in Fig. 4 provide strong evidence that the GRF complex does not contain Stat1␤.
Antibodies to Other STAT Proteins Do Not Supershift the pI␥RE⅐GRF Complex-Since recent reports have shown that numerous STATs and non-STAT proteins interact with Stat1 in nuclear transcription complex formation (19), we investigated whether GRF contained any other known STAT proteins by supershift EMSA. As shown in Fig. 4, both polyclonal and   FIG. 1. IFN␥ induces a specific pI␥RE-binding GRF complex in  HK cells in a time-dependent manner. A, EMSA using a 32 P-endlabeled pI␥RE probe incubated with 5 g of total nuclear extracts from HK that were either untreated (lane 1) or IFN␥-treated (lanes 2-5). A 10-fold molar excess of unlabeled competitor oligonucleotide was added as follows: distinct but irrelevant 30-bp double-stranded oligonucleotides (lanes 3 and 4) and pI␥RE (lane 5). B, EMSA using a 32 P-endlabeled pI␥RE probe incubated with 5 g of total cytoplasmic extracts from HK that were either untreated or treated with IFN␥ for specified times. Reactions with extracts from untreated cells are shown in lane 1, and reactions with extracts from cells treated with IFN␥ for varying times (in hours) are shown in lanes 2-6. monoclonal antibodies to Stat1␣ were able to supershift the pI␥RE⅐GRF complex. However, antibodies to Stat2, Stat3, Stat4, Stat5, and Stat6 did not interact with this complex (Fig.  5). Antibodies to p48, a component of the ISGF3 complex, also did not react with the GRF (data not shown).
GRF Contains a Distinctly Different DNA-Protein Complex-In order to investigate further whether the anti-Stat1␣ antibody supershifted GRF complex contained a form of Stat1 that could be competed by GBP⅐GAS or GBP⅐ISRE, we per-formed competition supershift EMSA. As seen in Fig. 6, the supershifted GRF when bound to pI␥RE was not competed by excess unlabeled double-stranded oligonucleotide corresponding to either the GAS or ISRE elements of the GBP gene. However, this complex was competed by excess of unlabeled pI␥RE and was also competed by double-stranded DNA corre- 1-7) or probes of pI␥RE mutants MUT 1 (lanes 8 and 9), MUT 2 (lanes 10 and 11), MUT 3 (lanes 12 and 13) or MUT 4 (lanes 14 and 15) were incubated with nuclear extracts from HK that were either untreated (lanes 1, 8, 10, 12, 14) or IFN␥-treated (lanes 2-7, 9, 11, 13, and 15).  7). B, a concentration-dependent competition EMSA using a 32 P-end-labeled pI␥RE probe incubated with 5 g of total nuclear extracts from HK that were either untreated (lane 1) or treated with IFN␥ (lanes 2-8). The molar ratio of unlabeled pI␥RE added as competitor to labeled pI␥RE is indicated in lanes 3-8. C, a concentration-dependent competition EMSA using a 32 P-end-labeled pI␥RE probe incubated with 5 g of total nuclear extracts from HK that were either untreated (lane 1) or treated with IFN␥ (lanes 2-20).   2-7). Polyclonal anti-Stat1␣ antibodies directed against the amino-terminal (lanes 4 and 5) or the carboxyl-terminal (lanes 6 and 7) were added to extracts prior to the addition of probe. A 10-fold molar excess of unlabeled pI␥RE was added (lanes 3, 5, and 7) as cold competitor. The pI␥RE⅐GRF complex and the supershifted complex are indicated. sponding to the IFNRE of the IRF-1 gene, an IFN␥RE that varies in sequence from pI␥RE by only a single nucleotide (see Table I).

FIG. 2. The pI␥RE mutants do not compete for the pI␥RE⅐GRF complex and do not form complexes with lysates from IFN␥-treated HK. A, 32 P-end-labeled pI␥RE probe (lanes
The pI␥RE⅐GRF Complex Displays a Distinct Mobility in EMSA, and pI␥RE Does Not Compete with the GAS⅐GAF Complex-Our data from the supershift EMSA using anti-Stat1␣ antibodies suggests that Stat1␣, or a Stat1␣-like protein, is part of the GRF that binds to the ICAM-1 pI␥RE, but both the GAS and ISRE elements failed to compete with pI␥RE for complex formation. Therefore, we investigated whether pI␥RE could cross-compete with GAS. In addition, because Stat1␣ had previously been shown to bind to the GAS element of several genes as a homodimer referred to as GAF (8), we also investigated whether the complexes formed with pI␥RE and GAS displayed any differences in EMSA. Using pI␥RE or GAS as probes with IFN␥-treated HK cell lysates, we observed striking differences in the mobility of complexes formed with pI␥RE and GAS when run in the same gel (Fig. 7). GRF displayed a distinctly slower mobility compared to the GAF complex. Furthermore, unlabeled pI␥RE at a 1000-fold molar excess failed to compete with GAS for the binding of GAF complex. However, a 1 to 1 molar ratio of unlabeled ISRE displayed significant competition with labeled GAS for binding GAF, while a 10-fold molar excess of unlabeled ISRE completely competed for GAF⅐GAS complex formation. More importantly, unlabeled GAS displayed a concentration-dependent competition for binding of GAF with a significant diminution of the labeled GAS⅐GAF complex when unlabeled GAS was added at only a 0.1 to 1 molar ratio. However, neither unlabeled GAS or ISRE, when added to labeled pI␥RE reactions, displayed any ability to compete for GRF complex formation even when added at a 1000-fold molar excess (Fig. 7). Interestingly, excess unlabeled IFNRE of the IRF-1 gene displayed a similar concentration-dependent competition for GRF complex formation to that displayed by unlabeled pI␥RE, suggesting similar binding affinities of GRF to these two IFN␥-responsive elements, as was the case using the Fc␥R1⅐GRR element as competitor for GRF binding in earlier experiments (Fig. 2C). These results indicate that GRF, which complexes with pI␥RE of ICAM-1, IFNRE of IRF-1, and GRR of Fc␥R1, is distinct from the classic GAF complex that binds to GAS.
GRF Displays Similarity with the DNA-Protein Complex Formed with the IFN␥RE of the Fc␥R1 Gene-In addition to the ability of IFNRE of the IRF-1 gene to compete for GRF binding to pI␥RE (Fig. 6 and 7), studies in our laboratory revealed that the IFN␥ response element pIRE of the ICSBP gene (29) also functioned well as a competitor for GRF binding (data not shown). We therefore investigated whether the complexes formed with pI␥RE and those formed with a representative element of those genes that contain IFN␥REs with sequences very similar to pI␥RE (see Table I and below) displayed similar or different mobilities in EMSA. Using labeled oligonucleotides of equal size (25 bp) containing either pI␥RE or the GRR of the Fc␥R1 gene (28) as probes, we compared EMSA mobilities of the DNA-protein complexes formed when these probes were FIG. 5. The pI␥RE⅐GRF complex is not recognized by antibodies to STAT proteins other than Stat1. EMSA using a 32 P-endlabeled pI␥RE probe incubated with 5 g of nuclear extracts from HK that were either untreated (lane 1) or IFN␥-treated (lanes 2-9). Anti-STAT antibodies were added as follows: polyclonal anti-Stat1␣

TABLE I Comparison of the sequences and GRF binding capacities of wild type and mutated IFN␥REs
Nucleotides in the IFN␥REs that are different from those of ICAM-1 pI␥RE are indicated by lowercase type. Mutated nucleotides in pI␥RE sequences designated as MUT 1 through MUT 4 are indicated in italics. The insertion mutation of 10 bp (ATGCATGCAT) in ICAM-1 MUT 3 is indicated as (10 bases), and the deletion mutation of the central G in ICAM-1 MUT 4 is indicated as an (X). The binding capacity of each sequence to bind GRF is indicated as follows: sequences that can bind to GRF, ϩ; sequences that do not bind to GRF, Ϫ.

TTTCCtGtAAA
Ϫ incubated with IFN␥-treated HK cell lysates. As seen in Fig. 8, the complexes that formed with pI␥RE and GRR displayed similar mobility. GRR also competed for GRF binding to pI␥RE, as seen in Fig. 2C, and unlabeled pI␥RE displayed a similar competition for GRF binding to GRR (data not shown). In addition, both complexes were supershifted in a similar manner with monoclonal antibodies to Stat1␣. Taken together, the data presented in these experiments indicate that pI␥RE of the ICAM-1 gene, GRR of the Fc␥R1 gene, IFNRE of the IRF-1 gene, and pIRE of the ICSBP gene form a subset of IFN␥ response elements that bind to a transactivating complex, ␥ response factor, that contains a Stat1␣like protein. This GRF complex clearly appears to be distinct from the homodimeric Stat1␣ complex that forms GAF and binds to the GAS element of other IFN␥-responsive genes, such as that characterized in the guanylate-binding protein gene. Comparison of the identified critical sequences of these GRFbinding and non-GRF-binding elements and their distinct differences, as well as the mutations used in the above studies, are shown in Table I. These data and comparisons indicate that a potential consensus sequence, TTTCNGNGAAA, is required for binding of GRF. They also indicate that elements such as GAS and ISRE, which do not contain the characteristic 11-bp palindrome sequence displayed by pI␥RE, bind to Stat1␣ containing complexes but not to the complex typified by the ␥ response factor, which binds to pI␥RE and similar elements identified in a subset of IFN␥-responsive genes. DISCUSSION We have characterized a specific DNA-binding complex, which we have termed GRF, that binds to the minimal IFN␥responsive element of the ICAM-1 gene, pI␥RE. Formation and binding of GRF is dependent on IFN␥-induced activation of pre-existing proteins, as demonstrated by the rapid activation and binding of GRF and by the activation of this complex in the presence of cycloheximide. 3 In addition, we have demonstrated that the pI␥RE sequence is necessary for binding GRF and that mutant pI␥RE sequences neither bind GRF nor compete with pI␥RE for the pI␥RE⅐GRF complex even when used at high molar ratios. Moreover, pI␥RE displays a high and specific binding affinity for GRF.
IFN␥ signaling involves activation of Stat1 by phosphorylation of a tyrosine residue in order to assemble active transcription-stimulating complexes (33,34). Treatment with PTPase 1-B, which specifically dephosphorylates tyrosine residues, has been shown to abrogate binding of these transcription complexes as shown by EMSA (34,35). Our data indicate that IFN␥-induced activation of GRF is tyrosine phosphorylationdependent as well.
It has been shown that Stat1␣ is activated by a  1-3) or pI␥RE probe (lanes 4 -6) was incubated with nuclear extracts from HK that were either untreated (lanes 1 and 4) or IFN␥-treated (lanes 2, 3, 5, and 6). Anti-Stat1␣ monoclonal antibody (amino-terminal) was added to reactions prior to addition of probe (lanes 3 and 6). The locations of pI␥RE⅐GRF and Fc␥R1⅐GRR-protein complexes (GRF) and anti-Stat1␣ supershifted complexes are indicated. other agonists, in addition to IFN␥ and IFN␣, based largely on reactivity of binding complexes with anti-Stat1␣ antibodies (19, 36 -38). Our supershift data using antibodies to the known STAT proteins suggest that Stat1␣, or a protein with antigenic similarities to Stat1␣, is a component of GRF. In addition, studies using polyclonal antibodies to either the amino-terminal or the carboxyl-terminal portions of Stat1␣ appear to exclude Stat1␤ as a possible component of the GRF complex. These data are in agreement with the recently reported GAFlike factor binding to the IFN␥RE of ICAM-1 gene (39) and the recent finding of Stat1␣ involvement in the complex by immunoblotting using anti-Stat1 antibodies (40). Previous reports indicate that Stat1␣ is a common component of the IFN-activated trans-acting complexes GAF and ISGF3. Stat1␣ binds to GAS as the homodimer GAF, and the multimeric ISGF3 complex, which binds to ISRE, contains Stat1␣ in addition to Stat2 and p48 (19). However, in the present studies, neither a doublestranded oligonucleotide corresponding to GBP⅐GAS nor GBP⅐ISRE used as unlabeled competitors competed with pI␥RE for the binding of GRF. Further, we have observed that double-stranded oligonucleotides corresponding to the IFN␥REs of the ISRE from the 6 -16 gene (27) and the GAS sequence (GCGGATCCTTTCCTGTAAAAGCTTGC) from the Ly6A/E gene (41) also fail to compete with pI␥RE for GRF complex formation. 4 Our studies demonstrate a lack of cross-competition between GAS and pI␥RE for formation of their respective DNA-protein complexes. However, ISRE from the GBP gene, which binds Stat1␣ as part of the heterotrimer in the ISGF3 complex and does not compete with pI␥RE for complex formation, clearly competes with GAS for the GAS⅐GAF complex. We have observed that both GAS and ISRE do not compete with pI␥RE even at a 1000-fold excess in our experiments using HK cell lysates. Our results are in agreement with the report that shows GBP⅐GAS does not compete with the ICAM-1 IFN␥RE in airway epithelial cells (20) and are in contrast to those obtained by others in MeL JuSo cells (42).
Consistent with our results, GRF that is supershifted by anti-Stat1␣ antibodies is not competed away by excess unlabeled GAS or ISRE from the GBP gene. On the other hand, excess unlabeled IFNRE from the IRF-1 gene displayed competition for the supershifted GRF. In fact, semi-palindromic IFN␥REs from other IFN␥-responsive genes, such as the IF-NRE of the IRF-1, pIRE of the ICSBP gene and GRR of Fc␥R1 gene, show somewhat closer sequence homology with that of pI␥RE than do the sequences of various GAS and ISRE elements as shown in Table I. It is precisely these elements that share greater sequence homology with pI␥RE that can function as competitors in EMSAs with pI␥RE for complex formation in IFN␥-induced HK cell lysates. Interestingly, these REs show differences in their sequences in nucleotides immediately flanking the G-C hinge of the pI␥RE. The mutations that rendered pI␥RE completely nonfunctional in vitro and in vivo in our study are either in the 5Ј or 3Ј end of the palindrome or the G-C hinge of the palindrome, and suggest binding specificity of GRF for these specific sequences. Comparison of the sequences of IFN␥REs of the IRF-1, ICSBP, and Fc␥R1 genes with the pI␥RE sequence of ICAM-1 and mutations of pI␥RE sequence reveals a potential consensus sequence, TTTCNGNGAAA, that is required for binding of GRF. Among the GAS and ISRE sequences of genes such as GBP, 6 -16, and Ly6A/E, the GAS sequence of Ly6A/E displays closest homology to this GRFbinding consensus sequence. However, the sequence of Ly6A/E diverges from the GRF-binding consensus sequence at two sites, the central G-C hinge and in the 3Ј half of the palindrome. We have been unable to demonstrate any competition for GRF binding with the LY6A/E⅐GAS. Whether this inability to bind GRF results from divergence of the nucleotide in the hinge or the 3Ј half of the palindrome has not yet been determined.
Finally, the significantly slower mobility of the pI␥RE⅐GRF complex compared to that of the GAS⅐GAF complex, using identical IFN␥-treated HK cell lysates with GAS and pI␥RE as separate probes in the same EMSA, cannot be accounted for by differences in the sizes of the radiolabeled oligonucleotides. In fact, the smaller pI␥RE probe (25 bp) formed a more slowly migrating DNA-protein complex when compared to the GAS⅐GAF complex bound by the larger (36 bp) GAS probe. In contrast, the IFN␥-activated DNA-protein complexes formed with either the Fc␥R1⅐GRR or the ICAM-1⅐pI␥RE clearly display similar mobility on EMSA, and both complexes supershift in a similar fashion with anti-Stat1␣ antibodies. The IFNRE of the IRF-1 gene and the pIRE of the ICSBP gene form a DNAprotein complex with IFN␥-treated HK cell lysates, which also displays a mobility similar to that of the pI␥RE⅐GRF complex in EMSA (data not shown).
Other investigators have recently shown that Fc␥R1⅐GRR does not compete with GBP⅐GAS in competition EMSA (43). Further, Fc␥R1⅐GRR was also shown to bind a Stat1␣-like protein that interacted with an additional 43-kDa protein in response to IFN␥-stimulation (35,44). The semi-palindromic IFN␥RE of the MIG gene, which displays significant sequence homology to ICAM-1⅐pI␥RE, has been reported to bind an IFN␥-activated trans-activating factor (␥RF-1) that is composed of at least two proteins of 95 and 130 kDa (45). Furthermore, ␥RF-1 was shown to exhibit differences in electrophoretic mobility distinct from GAF and to contain one or more subunits antigenically related to Stat1␣ (45).
These data thus indicate that pI␥RE represents a distinct subset of IFN␥REs found in a number of early response genes that mediate trans-activation in response to IFN␥ signaling through a DNA-binding protein complex (GRF) that is distinctly different from the previously characterized IFN␥-activated complex, GAF. While GRF certainly appears to contain a Stat1␣-like protein, identification and characterization of all of the components of this distinct trans-activating complex, their relationship to other STAT and non-STAT proteins, and the specific biochemical pathways involved in their activation will further elucidate the molecular mechanisms by which IFN␥ initiates differential responses at sites of localized inflammation.