Signaling by Covalent Heterodimers of Interferon-γ

Interferon-γ (IFN-γ) and its receptor complex are dimeric and bilaterally symmetric. We created mutants of IFN-γ that bind only one IFN-γR1 chain per dimer molecule (called a monovalent IFN-γ) to see if the interaction of IFN-γ with one-half of the receptor complex is sufficient for bioactivity. Mutating a receptor-binding sequence in either AB loop of a covalent dimer of IFN-γ yielded two monovalent IFN-γs, γm-γ and γ-γm, which cross-link to only a single soluble IFN-γR1 molecule in solution and on the cell surface. Monovalent IFN-γ competes fully with wild type IFN-γ for binding to U937 cells but only at a greater than 100-fold higher concentration than wild type IFN-γ. Monovalent IFN-γ had anti-vesicular stomatitis virus activity and antiproliferative activity, and it induced major histocompatibility complex class I and class II (HLA-DR) expression. In contrast, the maximal levels of activated Stat1α produced by monovalent IFN-γs after 15 min were never more than half of those produced by either wild type or covalent IFN-γs in human cell lines. These data indicate that while monovalent IFN-γ activates only one-half of a four-chain receptor complex, this is sufficient for Stat1α activation, major histocompatibility complex class I surface antigen induction, and antiviral and antiproliferative activities. Thus, while interaction with both halves of the receptor complex is required for high affinity binding of IFN-γ and efficient signal transduction, interaction with only one-half of the receptor complex is sufficient to initiate signal transduction.

IFN-␥ 1 is a bilaterally symmetric noncovalent dimer that binds two primary ligand-binding receptor chains (IFN-␥R1) in structurally equivalent positions (1,2). We define IFN-␥ as a divalent ligand because of this property. According to the current model of IFN-␥ signal transduction, after ligand binding and receptor complex activation (Fig. 1), the IFN-␥ receptor complex is composed of one IFN-␥ dimer, two IFN-␥R1 chains, two IFN-␥R2 chains, two Jak1 molecules, and two Jak2 molecules (3)(4)(5)(6)(7)(8). In this report, we address whether both halves of the receptor complex must be activated for signaling and whether a functional signaling receptor complex can be formed with only two receptor chains (one IFN-␥R1 and one IFN-␥R2 chain) instead of four chains. In order to address these questions, we designed and assayed a dimeric IFN-␥ molecule that can bind only one ligand binding receptor chain (a monovalent IFN-␥).
A tandem covalently linked dimer of IFN-␥ (9), designated ␥-␥, is an ideal ligand for studying the importance of ligand divalency for receptor activation because one of two receptor binding sites can be selectively eliminated to create a monovalent IFN-␥. The ␥-␥ was synthesized by linking two DNA sequences encoding monomers of wild type IFN-␥, a noncovalent IFN-␥ dimer designated ␥⅐␥, with DNA encoding a linker region from the IgA 1 molecule (9). This ␥-␥ had similar chromatographic properties as ␥⅐␥ and was conformationally similar to ␥⅐␥ as judged by 1 H NMR (9). In binding competition studies, ␥-␥ effectively competed with radiolabeled ␥⅐␥ for receptors on U937 cells, and its specific antiviral activity was 50 -65% that of ␥⅐␥ (9). With the two IFN-␥ monomers of ␥-␥ fused, it was possible to mutate each monomer segment separately to eliminate either one of the two receptor binding sites of ␥-␥. It was previously demonstrated that mutagenesis of residues 20 -25 of ␥⅐␥ within the AB loop (1,2) dramatically reduced the ability of ␥⅐␥ to bind to its receptor as judged by competition studies on U937 cells (10). These residues of the AB loop are part of a 3 10 helix that contacts IFN-␥R1 directly (2). The 1 H NMR spectrum of one such mutant, IFN-␥/A23E,D24E,N25K (IFN-␥/EEK; also designated ␥ m ⅐ ␥ m ), was nearly identical to that of ␥⅐␥, demonstrating that its overall structure is retained; however, it possessed less than 0.01% of the ability of ␥⅐␥ to compete for receptor binding to U937 cells (10). Consistent with its weak binding, this mutant possessed undetectable antiviral activity and stimulated HLA-DR␣ promoter activity only at over 100 nM protein (10). Thus, this mutation effectively eliminates the ability of IFN-␥ to bind to and activate its cellular receptor. Accordingly, we targeted these residues to inactivate each of the two receptor binding sites of ␥-␥ to create two monovalent ␥-␥ molecules. This report describes how these monovalent mutants of ␥-␥ were used to show that interaction of IFN-␥ with one-half of the receptor complex is sufficient to initiate signal transduction.

MATERIALS AND METHODS
Tissue Culture Media and Reagents-No antibiotics were used during propagation of cells. HEp-2 cells were grown in minimal essential medium supplemented with 10% (v/v) fetal bovine serum. HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. U937 cells were grown in RPMI medium supplemented with 1% fetal bovine serum in roller bottles at a density less than 1,000,000 cells/ml at 37°C, 5% CO 2 , 99% relative humidity. To prepare [ 125 I]IFN-␥, IFN-␥ (20 g) was incubated with 1 mCi of Bolton-Hunter reagent as described in the protocols of the supplier (NEN Life Science Products). Specific radioactivity was 32,000 cpm/ng of protein.
Construction of ␥ m -␥ and ␥-␥ m -The various ␥⅐␥ and ␥-␥ Ϫexpression plasmids were described previously (9,10). The pDI1 plasmid (9) was the source of the NH 2 -terminal half of the covalent dimer, and it encodes (in order from 5Ј to 3Ј) an EcoRI site, IFN-␥-(1-134), a HindIII site, the hinge region from IgA 1 , and ClaI and SalI restriction endonuclease sites. Plasmid pDI2 (9) was the source of the C-terminal half of the covalent dimer, and it contains a ClaI site in the 5Ј region encoding IFN-␥ and a SalI site after the IFN-␥ coding sequence. The p2106 plasmid encodes the IFN-␥/A23E,D24E,N25K mutant between EcoRI and SalI sites of the parental pTZ19U plasmid (10). The vector expressing ␥-␥ was constructed by ligating the pDI1 construct (retaining the NH 2 -terminal half of the dimer and the hinge region) digested with ClaI and SalI to the COOH-terminal half of the dimer excised from pDI2 with ClaI and SalI.
To insert the mutant IFN-␥ sequence into the first half of the ␥-␥ cDNA to yield CDEEK1, designated ␥ m -␥, the segment encoding the IFN-␥ mutant was excised from the p2106 plasmid with EcoRI and HindIII and was ligated in-frame into the pDI1 vector digested with EcoRI and HindIII. To insert the mutant IFN-␥ sequence into the second half of the ␥-␥ cDNA to yield CDEEK2, designated ␥-␥ m , a ClaI site replacing the EcoRI site at the 5Ј-end of IFN-␥ was created by polymerase chain reaction with plasmid p2106 as a template. A reverse primer based on the p2106 vector and the forward primer 5Ј-TCTAG-TATCGATGCAGGACCCATACGTGAAGGAA-3Ј were used for polymerase chain reaction with the p2106 vector as template. The boldface sequence encodes the polypeptide QDPYVKE, the amino acid sequence of IFN-␥- (2)(3)(4)(5)(6)(7)(8), and the underlined sequence represents a ClaI site in-frame with pDI1. The resulting polymerase chain reaction product was digested with ClaI and SalI and then ligated into the pDI1 vector digested with ClaI and SalI.
Protein Purification-Wild type IFN-␥ was purified according to Kung et al. (11). The ␥-␥ dimer was purified as described in Kung et al. (11) with some modifications. Bacteria expressing the IFN-␥ were solubilized by suspension in 7 M guanidinium hydrochloride, 5 mM EDTA, 40 mM NaCl, 50 mM Tris ⅐ HCl (pH 8.0). The lysate was cleared by centrifugation at 10,000 ϫ g for 10 min, and the supernatant was diluted dropwise in 10 volumes of 10 mM sodium borate (pH 8.3) with gentle stirring. Next, silica prewashed with PBS (NuGel-952AC, 40 -60 m diameter; Separation Industries, Metuchen, NJ) was added to this solution, and after binding in batch condition for 1 h, the IFN-␥ was eluted from a column packed with the above silica with 10 mM Na 3 BO 4 (pH 8.3), 1 M (NH 4 ) 2 SO 4 , 0.5 M N(CH 3 ) 4 Cl (buffer A) at 4 ml/min. Pooled fractions containing significant activity were applied directly without concentration to a hydrophobic FPLC column packed with phenyltoyopearl TSK 650S equilibrated with buffer A. All IFN-␥ activity appeared in the flow-through, which was concentrated with a Centiprep C-10 tube (10-kDa cut-off) and applied to a Superdex-75 FPLC column (Amersham Pharmacia Biotech) equilibrated in 25 mM ammonium acetate (pH 6.55), 10% glycerol. Elution was done with buffer A at 1.5 ml/min. The eluate containing the peak IFN-␥ activity was pooled and diluted with one volume of 25 mM ammonium acetate (pH 6.55), 6% glycerol and then applied to a semipreparative Mono S FPLC column and eluted at 1.5 ml/min with a linear gradient from 40 mM to 1 M NaCl in 25 mM NH 4 CH 3 COO (pH 6.55), 6% glycerol. IFN-␥ eluted between 0.35 and 0.4 M NaCl, and its purity was estimated to be over 95% by Coomassie Blue staining (data not shown).
Electrophoretic Mobility Shift Assay (EMSA)-The EMSA was performed essentially as described previously (14), with changes only in how the cells were lysed (15). Cells were plated in 12-well Falcon tissue culture dishes (catalog no. 3043), and were maintained at confluence for at least 2 days (1.3 ϫ 10 6 cells/well) prior to induction for 15 min with IFN-␥. Ligand was diluted to the working concentration at 30°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Twenty seconds prior to stop time, medium was removed, and cells were washed with 2 ml of PBS held at the same temperature. PBS was immediately removed, and at stop time (15 min), cells were scraped directly into 100 l of ice cold lysis buffer (0.5% Brij 96, 3 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, 1 mM sodium vanadate, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, and 150 mM NaCl). After a 30-min incubation on ice, 2.5 l of lysate was added to a mixture of 0.2 mg/ml poly(dI ⅐ dC), 1% Ficoll, 4 mM HEPES (pH 7.4), 30 g/ml bovine serum albumin, and approximately 0.4 ng of a 32 P-labeled oligonucleotide encoding the gamma-activating sequence (GAS) element from the IRF-1 promoter. This oligonucleotide was prepared from a self-annealing oligomer pair that was labeled with [␣-32 P]dCTP by filling in the four nucleotide overhangs with the Klenow fragment of DNA polymerase I (15,16) and has the final nucleotide sequence 5Ј-GATCGATTTCCCCGAAATCATG-3Ј with the Stat1␣ consensus binding site shown in boldface type. After incubation for 20 min at room temperature, samples were loaded onto a 5% native polyacrylamide gel and resolved by electrophoresis for 4 h at 450 V. Immediately afterward, the gel was dried under vacuum for 60 min at 80°C and autoradiographed, or its radioactivity was quantitated with a phosphor imager. Quantitation of radioactivity with a phosphor imager was performed with the GS-525 Molecular Imager system and the Molecular Analyst software supplied by Bio-Rad. All Stat1 responses are reported relative to that of a nonspecific band (17) whose intensity is proportional to the amount of cellular protein loaded onto the gel (Fig.  2). The presence of Stat1 was confirmed by supershift assay (data not shown). All data analysis and curve fitting were performed with SigmaPlot 3.02 or SigmaPlot 4.03 from Jandel Scientific.
MHC Assays-MHC class I surface antigen assays were performed as described (18). Briefly, cells were grown to 10 -20% confluence when induced with IFN-␥. Cells were then grown for 3 days in the continuous presence of ligand. Afterward, the cells (approximately 1 million) were washed with PBS, harvested with PBS supplemented with trypsin and 1 mM EDTA, and treated with 50 l of conditioned medium from a hybridoma secreting W6/32 murine monoclonal antibody (about 1 g) that reacts with an epitope of the HLA-B7 class I antigen (19). After a wash with PBS, fluorescein isothiocyanate-conjugated secondary antimouse antibody was added to the cells, and cells were analyzed by flow cytometry with a Coulter II flow cytometer.
HLA-DR␣ Reporter Gene Bioassay-The assay was performed as described previously (26). Briefly, the promoter region of the HLA-DR␣ gene (the HLADRA gene) was linked to the coding region of the growth hormone gene. This construct was transfected into HeLa cells, and clonal populations containing this construct were produced. Growth hormone was detected with an enzyme-linked immunosorbent assay (20).
Gel Filtration Chromatography-Superose 12 HR10/30 was packed into an FPLC column (24 ml, 10 ϫ 300 mm), and the column was equilibrated with 200 mM NaCl, 50 mM Tris ⅐ HCl (pH 8.0). Samples in a volume of 50 l in the same buffer were loaded onto the column through a loop, and elution proceeded at a flow rate of 0.4 ml/min with the same buffer. Fractions of 1.0 ml representing major expected peaks were collected manually. Standards for column calibration were obtained from Sigma (catalog no. MW-GF-200).
Binding Competition Studies-To perform the binding of the IFN-␥ species to U937 cells, the cells were concentrated by centrifugation and resuspended in 100-l aliquots at 1.25 ϫ 10 7 cells/ml in RPMI medium with no serum with 2 ng of 125 I-IFN-␥ (specific radioactivity of 32,000 cpm/ng) and various concentrations of unradiolabeled ligand (from 0.04 ng/ml to 40 g/ml). After incubation at 4°C for 2 h, cells were centrifuged through RPMI medium containing 5% sucrose, and the radioactivity in the pellet was counted in an LKB model 1272 ␥-counter to determine 125 I-IFN-␥ bound.
Solid-phase Receptor Binding Plate Assays-Purified Hu-IFN-␥R1 EC was absorbed to Nunc Maxisorb Immuno-Plates at a concentration of 0.4 g/100 l of PBS/well at 4°C overnight. The wells were drained and blocked with a solution of 5% nonfat milk in PBS for 2 h at ambient temperature. The wells were drained again, and 125 I-IFN-␥ was bound to the wells in 100 l of PBS containing 5% bovine serum albumin (w/v). After a 2-h incubation at ambient temperature, the wells were washed three times with PBS containing 5% Tween 20 (v/v). The 125 I-IFN-␥ bound was removed with 50 l of 0.2 N NaOH containing 1% SDS. Radioactivity was counted with an LKB model 1272 ␥-counter. Scatchard analyses were performed as described (21).
Cross-linking Procedures-Cross-linking to soluble receptors was carried out as described previously (13). Briefly, 16 M (12.5 g) recombinant soluble Hu-IFN-␥R1 EC produced from COS cells was incubated with 4 M (2.5 g) IFN-␥ in 50 l of PBS, and, after a 30-min incubation period, cross-linking was initiated by adding 5 l of 10 mM bis(sulfosuccinimidyl)suberate. After 30-min cross-linking, the reaction was stopped by the addition of 10 l of 1 M Tris ⅐ HCl, pH 7.0. Reaction products were analyzed by SDS-PAGE and visualized by Coomassie Blue staining.
Cross-linking to cell surface receptors was performed as described (3) with slight modification. Briefly, 125 I-labeled IFN-␥ species (2 nM ␥-␥ or 20 nM ␥-␥ m ) was added to 1 ϫ 10 6 U937 cells in 200 l of PBS containing 0.1% bovine serum albumin, followed by the addition of a noncleavable cross-linker EDC/sulfo-N-hydroxysuccinimide (10 mM each; EDC added immediately before sulfo-N-hydroxysuccinimide) and further incubation for 60 min at ambient temperature. The reaction was stopped by the addition of 0.1 volume of 1 M glycine. Cells were washed once with PBS and then lysed as described (3). The whole cell lysate was analyzed by SDS-polyacrylamide gel electrophoresis, and the cross-linked products were visualized by autoradiography.
Antiviral Assays-Antiviral assays were performed essentially as previously reported (22). HeLa cells (35,000 per well) were plated in 96-well microtiter plates containing 12 2-fold serial dilutions of IFN-␥. After 24 h, vesicular stomatitis virus was added at 5000 plaque-forming units/well. When cells in viral control wells containing no interferon were lysed (usually about 48 h), plates were drained and stained with crystal violet to visualize live cells.

Stoichiometry of Ligand Receptor Interactions-
The stoichiometry of the covalent dimer of IFN-␥ (␥-␥) with the soluble IFN-␥R1 extracellular domain (IFN-␥R1 EC ) was examined by gel filtration to ensure that the ␥-␥ could bind two soluble IFN-␥R1 (IFN-␥R1 EC ) molecules. It was previously demonstrated that wild type IFN-␥ (␥⅐␥) (33-34 kDa) can bind to two IFN-␥R1 EC s (40 kDa) or two cell surface receptors simultaneously (3,12,13). Thus, a soluble ternary complex would have a molecular mass of 114 kDa, and a binary complex would have a molecular mass of 74 kDa. As shown in Fig. 3, both ␥-␥ and ␥⅐␥ form receptor complexes greater than 100 and 130 kDa apparent molecular mass, respectively, independent of the ratio of ligand to receptor. Our results with ␥⅐␥ agree well with the gel filtration data reported for ␥⅐␥ (3) and were repeated here as a positive control. Western blotting confirmed that fractions containing the higher molecular weight complex were composed of both ligand and receptor, while fractions containing the lower molecular weight complex were composed almost entirely of the component in excess, either ligand or receptor (data not shown). Small traces of the high molecular weight complex are present in the fractions containing the lower molecular weight component because the two peaks were not completely resolved. The ␥-␥ ⅐ IFN-␥R1 EC complex may have a lower apparent molecular weight because either its complex may be less stable than the ␥⅐␥ ⅐ IFN-␥R1 EC complex during gel filtration or the hinge region of ␥-␥ may alter the structure of the ternary complex.
Although the (␥-␥) ⅐ IFN-␥R1 EC and (␥⅐␥) ⅐ IFN-␥R1 EC complexes were consistently above 100 and 130 kDa, respectively, variability in the physical structures or stability of proteins and protein complexes make accurate molecular weight and stoichiometric determinations from a size exclusion column ambiguous. To confirm the stoichiometry of each ligand-receptor complex, we cross-linked the ligand-soluble receptor complexes, analyzed the products by SDS-polyacrylamide gel electrophoresis, and estimated their aggregate apparent molecular weights. As shown in Fig. 4A L 2 R 2 , composed of one ligand dimer (33-34 kDa; ␥⅐␥, ␥-␥, ␥ m -␥, or ␥-␥ m (L 2 )) and either one IFN-␥R1 EC (36 -40 kDa (R)) or two IFN-␥R1 EC s (72-80 kDa (R 2 )) (22). Although ␥⅐␥ also formed complexes of similar sizes, most of the cross-linked products appeared at about 52 and 90 kDa so that the bands at 70 -75 kDa and 105-115 kDa appeared less intense (Fig. 4A, left small  panel). Because ␥⅐␥ is a noncovalent dimer, the 52-and 90-kDa complexes represent a cross-link of one or two receptors to a single chain of ␥⅐␥, which is 17 kDa shorter than the covalent single chain of ␥-␥, ␥ m -␥, or ␥-␥ m . This is consistent with the previous demonstration that ␥⅐␥ formed similar molecular weight cross-link patterns with IFN-␥R1 EC (3,13). Thus, ␥-␥, like ␥⅐␥, is a divalent ligand. The ligand binding site mutants ␥ m -␥ and ␥-␥ m form complexes of only 70 -75 kDa, with no significant amount of the 105-115-kDa complex, confirming that ␥ m -␥ and ␥-␥ m are monovalent ligands.
Cross-linking-To determine whether ␥-␥ m is monovalent on the cell surface as well as in solution, we cross-linked ␥-␥ and ␥-␥ m to the IFN-␥ receptor chains on the surface of U937 cells. It was already previously shown that ␥⅐␥ was able to cross-link to two IFN-␥R1s simultaneously as well as to IFN-␥R2 on the cell surface (3,4,6). The ␥-␥ efficiently cross-linked to two molecules of IFN-␥R1 (Fig. 4B, lane 1), demonstrating that ␥-␥ behaved as a divalent ligand on the cell surface. ␥-␥ m could cross-link efficiently to only one IFN-␥R1 (lane 3), demonstrat-ing that ␥-␥ m functions as a monovalent ligand both in solution as well as on the cell surface. Both ligands also appear in cross-linked bands of about 20 kDa below the L 2 R and L 2 R 2 bands. We believe that this is a cross-link between either L 2 or L 2 R and IFN-␥R2, since their molecular weights correspond to those of cross-linked bands predicted between the ligand dimer (34 kDa), IFN-␥R1 (70 kDa), and IFN-␥R2 (60 kDa), which has previously been reported to be present in cross-linked complexes (4,6). The cross-linking results with ␥ m -␥ are not displayed because the cross-linking was too low compared with the control with excess unlabeled ligand to be deemed significant.
Binding of Various IFN-␥ Molecules to Cells and Soluble Receptors-To estimate the relative affinities of monovalent IFN-␥ for the cell surface, competition studies of each ligand against 125 I-␥⅐␥ were performed on U937 cells. Both ␥⅐␥ and ␥-␥ are approximately equivalent in competing with 125 I-␥⅐␥ (Fig.  5). Both ␥ m -␥ and ␥-␥ m can fully compete with ␥⅐␥ for binding to U937 cells; however, the amounts of ␥ m -␥ and ␥-␥ m necessary for 50% competition (9 and 3.5 nM, respectively) were over 2 orders of magnitude higher than that of ␥⅐␥ (20 pM) and ␥-␥ (30 pM). This result suggests that the monovalent IFN-␥ ligands can compete with ␥⅐␥ but that their affinity for the cell surface receptors is much lower. It appears that IFN-␥ must be divalent for high affinity cell surface binding. If IFN-␥ must be divalent to bind to the cell surface with high affinity, then the residual competitive ability of ␥ m -␥ and ␥-␥ m could result from a reconstitution of a divalent wild type IFN-␥ ligand from fragments resulting from degradation of the monovalent IFN-␥. Alternatively, this reconstituted divalent IFN-␥ could be produced by intermolecular association of two intact covalent heterodimers to produce a covalent tetramer acting as a divalent IFN-␥. To examine the possibility that a reconstituted divalent IFN-␥ is the active species in ␥-␥ m preparations, binding studies of ␥-␥ m and ␥-␥ were performed with IFN-␥R1 EC -coated plates and were analyzed by the method of Scatchard (21). Fig. 6 shows the results of the Scatchard plots of the two ligands. The divalent ␥-␥ possesses two binding affinities for the IFN-␥R1 EC immobilized to the microtiter plates. Five to ten percent of the binding sites on the plates have an affinity for ␥-␥ of 150 pM, identical to the affinity found for ␥⅐␥ on U937 cells by Rashidbaigi et al. (24). We believe that this affinity represents the true affinity of ␥-␥ for cellular receptors present on the U937 cell surface, because in the cellular competition assays all of the prebound radiolabeled ␥⅐␥ was displaced with less than 1 nM ␥-␥ (Fig. 5). The remaining binding sites on the plate have an affinity for ␥-␥ of about 10 nM. ␥-␥ m has only the 10 nM affinity for the immobilized receptors, with no trace of the high affinity site. This affinity also resembles that observed in the cellular competition assay. Because no high affinity binding was observed with ␥-␥ m , we conclude that there is no reconstituted ␥⅐␥ divalent ligand in the ␥-␥ m preparation and that all cellular receptor binding activity is a direct result of binding of the intact ␥-␥ m to the receptors. Consistent with the observation that high affinity binding required ligand divalency, both ␥-␥ and ␥-␥ m possess the same low affinity constant for immobilized IFN-␥R1 EC . The origin of the nanomolar affinity binding for ␥-␥ may be related to the lack of specific orientation of IFN-␥R1 EC molecules on the plate. Because the recep- The arrows to the left indicate the complexes with the indicated stoichiometry of cell surface IFN-␥R1 (R) and ligand dimer (L 2 ). Estimated molecular masses of labeled complexes are 210, 120, and 32 kDa from top to bottom, respectively. L 2 R 2 is a ternary complex of IFN-␥ and two IFN-␥R1 chains; L 2 R is a binary complex of IFN-␥ and one IFN-␥R1 chain; L 2 is free IFN-␥ (for details, see above).
tors on the surface of the microtiter plate are probably immobilized randomly, only a small fraction of the receptors may be oriented to allow divalent ligand binding to two IFN-␥R1 EC molecules. Therefore, even with divalent ligand preparations, the predominant binding may occur to single immobilized IFN-␥R1 EC molecules. Thus, the 10 nM affinity of ␥-␥ probably represents the affinity of ␥-␥ for a single IFN-␥R1 EC .
The above results demonstrate that the heterodimeric ␥-␥s (␥-␥ m and ␥ m -␥) are monovalent ligands with severely reduced affinity for cellular receptors. No evidence for divalent IFN-␥ present in the ␥ m -␥ and ␥-␥ m preparations was observed.
Antiviral, MHC Class I Antigen Induction, HLA-DR␣, and Stat1␣ Assays-To test the hypothesis that a single receptor activation event is sufficient to activate IFN-␥ signal transduction, we employed several bioassays. Both divalent IFN-␥ (␥ ⅐␥ and ␥-␥) and monovalent IFN-␥ (␥ m -␥ and ␥-␥ m ) possessed antiviral activity (Fig. 7). The cytopathic effect end point (ED 50 ) with HeLa cells was 3 pM (1 unit/ml) for ␥⅐␥ and 5 pM for ␥-␥, as previously reported by Lunn et al. (9) (Table I). The ED 50 was 36 nM for ␥ m -␥ and 390 pM for ␥-␥ m . At much higher concentrations of IFN-␥, both monovalent as well as divalent IFN-␥ began to exhibit antiproliferative effects on HeLa cells, as seen in the leftmost wells in Fig. 7. Analysis of the slowly proliferating cells by light microscopy confirmed that all cells appeared uninfected and alive.
MHC class I surface antigen expression was induced by 3 days of treatment of HEp-2 cells by monovalent and divalent IFN-␥ (Fig. 8). Upon induction by 300 nM ␥ m -␥ and 90 nM ␥-␥ m , the MHC class I surface antigen expression was equivalent to that induced by 3 nM ␥-␥ or 3 nM ␥⅐␥. We also assayed the ability of the four IFN-␥s to activate the HLA-DR␣ promoter controlling growth hormone (GH) expression in HeLa cells. Both monovalent and divalent IFN-␥ were able to maximally induce GH secretion in response to IFN-␥ treatment (Fig. 9). Consistent with previous results, much higher concentrations of monovalent IFN-␥ (1.5 nM ␥ m -␥ and 40 pM ␥-␥ m ) were necessary for half-maximal increases in GH secretion than required for divalent IFN-␥ (0.3 pM ␥⅐␥ and 3.5 pM ␥-␥). As a control, the IFN-␥ mutant that contains the EEK mutation (␥ m ⅐ ␥ m ) on both chains was simultaneously assayed. It possessed much less potency than either monovalent ligand, with an ED 50 of 100 nM for HLA-DR-driven GH secretion (Fig. 9) in agreement with previous results (10).
Stat1␣ activation is an early event of IFN-␥ signal transduc-

FIG. 5. Competition of various ligands with IFN-␥.
The ␥⅐␥ was radioiodinated and bound to U937 cells, and the different IFN-␥s were assayed for their ability to compete with radiolabeled ␥⅐␥ for cellular receptors on the surface of U937 cells as a function of ligand concentration. Unfilled circles, ␥⅐␥; filled circles, ␥-␥; unfilled triangles, ␥-␥ m ; filled triangles, ␥ m -␥. Nonspecific binding in the presence of excess IFN-␥ was less than 1% of total binding and was subtracted from the total binding to yield specific binding for the data of the figure.
FIG. 6. Binding of ␥-␥ and ␥-␥ m to immobilized receptors. The Scatchard analysis (21) of the binding data is shown. The binding of 125 I-labeled ␥-␥ and ␥-␥ m to microtiter plates containing immobilized IFN-␥R1 EC as a function of the ligand concentrations is shown in the inset. Filled circles, ␥-␥; unfilled triangles, ␥-␥ m . Nonspecific binding was about 10% of the total binding and was subtracted from the total binding to obtain specific binding for the data in the figure. tion necessary for both antiviral activity and induction of various genes (25)(26)(27). The abilities of the IFN-␥ variants to activate Stat1␣ in human HeLa cells after 15 min are shown in Fig. 10. The results are similar with other human cell lines (Table I). Stat1␣ is activated by both monovalent and divalent IFN-␥, and the extent of activation increases with increasing IFN-␥ concentration. Approximately 6 pM ␥⅐␥ yields a halfmaximal activation level of Stat1␣ in HeLa cells. Although ␥-␥ activates as much Stat1␣ as ␥⅐␥ at its optimal levels, the EC 50 of 300 pM for ␥-␥ in this assay was 50-fold higher. Monovalent IFN-␥ activated Stat1␣ only at much higher concentrations. Half-maximal Stat1␣ activation was observed at 80 nM ␥ m -␥

TABLE I Activities of the various Hu-IFN-␥s
Each IFN-␥ was assayed for various biological activities. Concentrations in pM required to elicit half-maximal response for the various activities are tabulated along with the cell lines used for each assay. For Stat1␣ data, the first number represents the EC 50 of the IFN-␥, while the number in parentheses represents the maximal level of phosphorylated Stat1␣ in 15 min relative to that of ␥ ⅐ ␥, arbitrarily set as 100. MHC class I data are presented as a pair. The first number is the picomolar concentration of IFN-␥ required to induce maximal MHC class I antigen expression. The second number is the ratio of fluorescence of cells binding W6/32:fluorescein isothiocyanate-conjugated anti-murine IgG complexes after IFN-␥ treatment relative to fluorescence of untreated cells.

Activity
Cell line  In other human cell lines tested, the maximal levels of Stat1␣ activated by ␥ m -␥ or ␥-␥ m were never more than about half that of ␥⅐␥ or ␥-␥ (Table I). These results are consistent with the hypothesis that the Jak/STAT pathway can be activated by monovalent IFN-␥ binding to and activating one-half of the receptor complex. Thus, it would seem that while only a single side of the IFN-␥ receptor complex needs to be activated to elicit IFN-␥ bioactivity, ligand divalency is required for a maximal Stat1␣ response as well as for high affinity receptor binding. DISCUSSION Our results show that ␥-␥, like ␥⅐␥, was capable of binding two soluble receptors as demonstrated by gel filtration chromatography and by covalent cross-linking to two soluble receptors and to two cell surface IFN-␥R1 chains (Figs. 3 and 4, A and B). The Ala-Asp-Asn sequence in the AB loop at the NH 2 terminus of ␥-␥ resides in the portion of ␥-␥ that closely resembles either half of ␥⅐␥. Mutation of this sequence to create ␥ m -␥ results in a molecule that can only signal with the half that contains AB loop near the IgA 2 hinge (Fig. 11). Mutation of the analogous sequence in the COOH-terminal half of the primary structure of ␥-␥ to create ␥-␥ m eliminates receptor binding in the hinged region. ␥-␥ m can bind to receptors only with the half that resembles either half of ␥⅐␥ (Fig. 11). Each monovalent IFN-␥ can cross-link to only one soluble receptor, and ␥-␥ m can crosslink efficiently to only one cell surface IFN-␥R1 (Fig. 4, A and  B), demonstrating that this mutation in ␥-␥ created monovalent IFN-␥. The hinge region, perhaps by restricting the optimal conformations of the termini of IFN-␥, impairs both the binding and the biological activity of IFN-␥, as seen in comparing the binding and activity of ␥-␥ with those of ␥⅐␥ and by comparing the binding and activity of ␥ m -␥ with those of ␥-␥ m ( Table I). The relative placement of the mutations in the AB loop with respect to the hinge appears to distinguish the ␥ m -␥ and ␥-␥ m variants. Both ␥ m -␥ and ␥-␥ (whose receptor binding site is close to the hinge) are less efficient at competition and have less biological activity than do ␥-␥ m and ␥⅐␥, respectively ( Table I).
The monovalent IFN-␥s had considerably decreased affinity for the cell surface receptor as shown in competition studies and solid-phase binding assays (Figs. 5 and 6). The nanomolar affinity of ␥-␥ m and ␥ m -␥ does not appear to arise from trace contaminants of divalent IFN-␥ because 1) divalent IFN-␥ had two saturable specific binding sites in the solid-phase receptor binding assay, while ␥-␥ m had only a single binding site (Fig. 6); 2) divalent IFN-␥ formed a 105-115-kDa complex when crosslinked with the soluble IFN-␥R1 EC chain in addition to the 70 -75-kDa complex, while ␥-␥ m and ␥ m -␥ formed only the 70 -75-kDa complex (Fig. 4A); and 3) monovalent IFN-␥ never activated as much Stat1␣ in 15 min as did ␥⅐␥ or ␥-␥ in any human cell line tested (Fig. 10A, Table I). If any activity in the ␥ m -␥ or ␥-␥ m preparations resulted from the presence of a divalent IFN-␥ contamination, the maximum Stat1␣ activation ultimately would be the same for ␥ m -␥ and ␥-␥ m as for ␥⅐␥ or ␥-␥. However, this was not seen (Fig. 10, Table I). All of these data led us to conclude that there is no detectable trace of divalent ligand in the monovalent IFN-␥ preparations and that all residual activity from the monovalent ligand preparations arose from the binding of monovalent ligands to low affinity sites on the cell surface.
Both ␥-␥ m and ␥-␥ have the same 10 nM affinity for the immobilized IFN-␥R1 EC (Fig. 6). If the 10 nM affinity seen with ␥-␥ represents an intermediate binary complex before formation of the ternary ␥-␥(IFN-␥R1) 2 complex (Figs. 3 and 4A), divalency of IFN-␥ is apparently required for high affinity binding. Alternatively, the increased affinity with divalent ligand binding may be mediated by avidity. Avidity is a term often used to describe an increase in apparent affinity of multivalent antibodies relative to that of monovalent Fab frag-ments due to multiple interactions, each having identical affinity with an antigen. Significantly, the gel filtration experiments show that although a ternary complex between ␥⅐␥ or ␥-␥ and IFN-␥R1 EC is spontaneously formed, even at low IFN-␥R1 EC :ligand ratios, the apparent molecular weight of the ligand-receptor complex gradually increases as more receptor is titrated into the reaction. This is consistent with a rapid binary:ternary equilibrium between ␥-␥ or ␥⅐␥ and soluble IFN-␥R1 EC . Soluble IFN-␥R1 EC can displace ␥⅐␥ from the IFN-␥ receptor complex on the cell surface and inhibit signal transduction, indicating that this equilibrium does indeed occur on the cell surface as well as in solution (3). Since a single receptor activation event is sufficient to trigger IFN-␥ signaling, the use of soluble IFN-␥ receptors as IFN-␥ signaling antagonists may not be as effective as growth hormone (28), because IFN-␥ bound to one soluble IFN-␥R1 and one cell surface IFN-␥R1 would signal according to our results.
Remarkably, monovalent IFN-␥ has full biological activity. This contrasts with results obtained with growth hormone where mutation of the second receptor binding site creates an antagonist of the receptor complex (28). Randal and Kossiakoff (29) noted that their monovalent IFN-␥ was also active in signal transduction but provided no supporting biological data. Our monovalent IFN-␥ exhibited antiviral and antiproliferative activity (Table I, Fig. 7). It activated the HLA-DR␣ promoter and induced MHC class I surface antigen expression (Table I, Figs. 8 and 9). In accord with these activities, our FIG. 11. Schematic diagrams of IFN-␥'s. Each IFN-␥ used in this study is depicted schematically. The region of the EEK mutation is represented by a filled square. N, the NH 2 terminus of the molecule; C, the COOH terminus of the molecule. The black loop represents the IgA 2 hinge region. monovalent IFN-␥s also activated the Jak/STAT pathway in several human cell lines, however only to about half the extent as divalent IFN-␥ (Fig. 10, Table I). Because monovalent IFN-␥ can bind and cross-link to only one IFN-␥R1 molecule, these data indicate that half of the receptors normally activated with divalent IFN-␥ cannot be activated with monovalent IFN-␥ and suggest that the active signaling complex consists of one IFN-␥R1 bound and activated by monovalent IFN-␥ while the second IFN-␥R1 is not activated and is not accessible to another molecule of monovalent IFN-␥ (Fig. 12). This is consistent with the hypothesis that the endogenous receptor complex is composed of two IFN-␥R1 chains whether the ligand is divalent or monovalent and that the IFN-␥ receptor complex is preformed. Therefore, monovalent IFN-␥ activates only one side of the receptor complex, leaving the other side of the complex unactivated. Although only one side of the IFN-␥ receptor complex is signaling with monovalent IFN-␥, it is able to confer full IFN-␥ activity, suggesting that each half of the receptor complex acts independently and that one-sided signaling is sufficient for IFN-␥ signal transduction. A monomeric fragment of human IFN-␥, consisting of amino acids 95-134, was reported to bind to amino acids 253-287 within the cytoplasmic domain of human IFN-␥R1 with a dissociation constant of 37 nM in vitro (30). This monomeric peptide expressed in cells induced MHC class II surface antigen expression and conferred anti-vesicular stomatitis virus protection without species specificity (31). These studies also infer that a single receptor activation event can initiate IFN-␥ signaling; however, the stoichiometry of peptide to receptor complex was not reported. Thus it cannot be excluded that two peptide molecules per receptor complex may be required to initiate signaling.
The concentrations of ␥⅐␥ and ␥-␥ required for binding competition, antiviral, and antiproliferative activities differ by only a factor of 2 (Fig. 7, Table I), suggesting a minimal influence from the IgA 1 linker segment on these activities. In contrast, 10 -50-fold higher levels of ␥-␥ than ␥⅐␥ are required for full activation of the Jak/STAT pathway or activation of the HLA-DR␣ promoter. Moreover, both ␥-␥ m and ␥-␥ induce anti-VSV activity at much lower concentrations than those needed to observe significant levels of Stat1 activation, a phenomenon we have observed with several type I interferons 2 (Figs. 8 and 10, Table I). This surprising result supports the hypothesis that induction of antiviral and antiproliferative activity could involve many different components in addition to Stat1␣ activation, which is necessary but not sufficient to elicit these activities (26,27,32,33). Alternatively, a pathway that does not require Stat1␣ activation may mediate some of these uncharacterized pathways. Therefore, a more complex IFN-␥ signaling network is likely to exist in which only a fraction of the components have yet been identified. Our studies provide a basis to begin to elucidate these pathways and their components.
In summary, monovalent IFN-␥ that only activates one receptor chain is active in signal transduction, and activation of a single side of the IFN-␥ receptor complex is sufficient to yield full IFN-␥ bioactivity. However, interaction of ligand with both sides of the complex is required for high affinity activity of IFN-␥ and maximal, efficient activation of the IFN-␥ receptor complex.