The Stability of the Ternary Interferon-Receptor Complex Rather than the Affinity to the Individual Subunits Dictates Differential Biological Activities*

Type I interferons (IFNs) signal for their diverse biological effects by binding a common receptor on target cells, composed of the two transmembrane IFNAR1 and IFNAR2 proteins. We have previously differentially enhanced the antiproliferative activity of IFN by increasing the weak binding affinity of IFN to IFNAR1. In this study, we further explored the affinity interdependencies between the two receptor subunits and the role of IFNAR1 in differential IFN activity. For this purpose, we generated a panel of mutations targeting the IFNAR2 binding site on the background of the IFNα2 YNS mutant, which increases the affinity to IFNAR1 by 60-fold, resulting in IFNAR2-to-IFNAR1 binding affinity ratios ranging from 1000:1 to 1:1000. Both the antiproliferative and antiviral potencies of the interferon mutants clearly correlated to the in situ binding IC50 values, independently of the relative contributions of the individual receptors, thus relating to the integral lifetime of the complex. However, the antiproliferative potency correlated throughout the entire range of affinities, as well as with prolonged IFNAR1 receptor down-regulation, whereas the antiviral potency reached a maximum at binding affinities equivalent to that of wild-type IFNα2. Our data suggest that (i) the specific activity of interferon is related to the ternary complex binding affinity and not to affinity toward individual receptor components and (ii) although the antiviral pathway is strongly dependent on pSTAT1 activity, the cytostatic effect requires additional mechanisms that may involve IFNAR1 down-regulation. This differential interferon response is ultimately mediated through distinct gene expression profiling.

Type I interferons (IFNs) signal for their diverse biological effects by binding a common receptor on target cells, composed of the two transmembrane IFNAR1 and IFNAR2 proteins. We have previously differentially enhanced the antiproliferative activity of IFN by increasing the weak binding affinity of IFN to IFNAR1. In this study, we further explored the affinity interdependencies between the two receptor subunits and the role of IFNAR1 in differential IFN activity. For this purpose, we generated a panel of mutations targeting the IFNAR2 binding site on the background of the IFN␣2 YNS mutant, which increases the affinity to IFNAR1 by 60-fold, resulting in IFNAR2-to-IFNAR1 binding affinity ratios ranging from 1000:1 to 1:1000. Both the antiproliferative and antiviral potencies of the interferon mutants clearly correlated to the in situ binding IC 50 values, independently of the relative contributions of the individual receptors, thus relating to the integral lifetime of the complex. However, the antiproliferative potency correlated throughout the entire range of affinities, as well as with prolonged IFNAR1 receptor down-regulation, whereas the antiviral potency reached a maximum at binding affinities equivalent to that of wild-type IFN␣2. Our data suggest that (i) the specific activity of interferon is related to the ternary complex binding affinity and not to affinity toward individual receptor components and (ii) although the antiviral pathway is strongly dependent on pSTAT1 activity, the cytostatic effect requires additional mechanisms that may involve IFNAR1 down-regulation. This differential interferon response is ultimately mediated through distinct gene expression profiling.
Type I interferons (IFNs) 2 form a class of cytokines capable of mediating antiviral, growth-inhibitory and immunoregulatory activities (1)(2)(3). Comprising about 18 members in humans (4), all IFNs induce their biological activities through binding to the same receptor complex, composed of the two transmembrane proteins IFNAR1 and IFNAR2 (5). The signal is then transduced mainly through the JAK-STAT pathway, in which the tyrosine kinases Tyk2 and Jak1, which are constitutively associated with the IFNAR1 and IFNAR2 subunits, are activated by phosphorylation upon IFN-receptor binding (6,7). Subsequently, they phosphorylate STAT proteins that dimerize or form ternary complexes with DNA-binding proteins that translocate to the nucleus to promote the expression of IFN-stimulated genes. The IFNAR1 and IFNAR2 receptor subunits are expressed in virtually every nucleated cell (6), yet the way in which IFNs exert specific activities in different physiological contexts is still obscure. Despite common biological activities, differences observed for various IFNs (8 -11) seem to indicate that at least not all IFN subtypes are redundant. Different potencies of specific IFNs are known to be determined to a large extent by their affinities toward the IFNAR1 and IFNAR2 subunits, with only IFN␤ standing out for its higher affinity toward IFNAR1 (11). Significant differences are observed in the ability of the two subunits to bind IFNs, as in the case of IFN␣2, which binds the IFNAR1 and IFNAR2 subunits with micromolar and nanomolar affinities, respectively. Nevertheless, both components are necessary to induce the IFN signal, as demonstrated by the use of neutralizing antibodies against the extracellular part of IFNAR1 (12) and by a cell line lacking IFNAR2 (13), both resulting in complete lack of responsiveness to both IFN␣ and IFN␤.
We have recently described an optimized IFN␣2 mutant (YNS), which binds IFNAR1 with ϳ60-fold higher affinity compared with the wild type (K D ϭ 30 nM) while keeping its affinity toward IFNAR2 unaltered (K D ϭ 1.5 nM) (14). This is feasible because the binding sites on interferon toward both receptors are located on opposite surfaces and are independent of each other (15)(16)(17)(18). In WISH cells, the YNS mutant exhibited a 100fold stronger antiproliferative potency compared with its wild type counterpart, accompanied by only a slight increase in its antiviral potency, suggesting that IFNAR1 might play a unique role in the induction of IFN-promoted antiproliferative signals and account for differential signaling within this cytokine family (11,19,20). Here, we test this hypothesis by analyzing a set of IFN␣2 mutants, in which the optimized IFNAR1 binding was combined with a reduction in affinity toward IFNAR2 to various degrees. Our results indicate an additive contribution of both IFNAR1 and IFNAR2 binding to antiproliferative potency along the whole range of mutations, whereas antiviral potency reaches a maximum already at low combined affinities. The differential biological effects, which relate to qualitatively different gene expression patterns, are distinctly related to receptor down-regulation and pSTAT1, suggesting that the antiproliferative and antiviral effects are driven by different pathways.

EXPERIMENTAL PROCEDURES
Protein Mutagenesis-Site-directed mutagenesis of IFN␣2 was carried out by primer extension using bacteriophage T7 DNA polymerase on expression vector pT7T318U containing the IFN␣2 gene, which was used as a single-stranded DNA template following subsequent transfection into and recombinant protein production in CJ236 cells (21).
Protein Expression and Purification-IFN␣2 and IFNAR2-EC were expressed in E. coli Rosetta strain and purified by ion exchange and size exclusion chromatography (22). Protein concentrations were determined both by analytical gel filtration chromatography and from the absorbance at 280 nm with ⑀ 280 ϭ 18,500 cm Ϫ1 M Ϫ1 for IFN␣2 and ⑀ 280 ϭ 26,500 cm Ϫ1 M Ϫ1 for IFNAR2-EC. IFNAR1-EC was expressed in Sf9 insect cells and purified as described before (23).
In Vitro Binding Assays-Binding affinities of IFN␣2 toward IFNAR1-EC or IFNAR2-EC were measured using the ProteOn XPR36 Protein Interaction Array system (Bio-Rad), based on surface plasmon resonance technology. Phosphate-buffered saline was used as running buffer at a flow rate of 30 l/min. For immobilization of the IFNAR1 receptor subunit, a Neutr-Avidin-covered sensor chip (NLC; Bio-Rad) was bound with 150 nM biotin-conjugated tris-nitrilotriacetic acid (24). The IFNAR1-EC subunit was then injected to separate channels in duplicates, at 100 and 200 nM, respectively, to bind the trisnitrilotriacetic acid on the chip surface via their His tags. The IFNAR2 receptor was immobilized as detailed in Ref. 14. The tested interferons were then injected perpendicular to ligands at six different concentrations within a range of 16.5-4,000 nM for IFNAR1 binding and 6.25-200 nM for IFNAR2 binding. Data were analyzed using BIAeval 4.1 software, using the standard Langmuir models for fitting kinetic data. Dissociation constants K D were determined from the rate constants according to Equation 1, or from the equilibrium response at six different analyte concentrations, fitted to the mass-action equation (22). Binding Assays on Artificial Membranes-Binding of IFN␣2 to IFNAR1-EC and IFNAR2-EC tethered onto artificial membranes was monitored in real time by simultaneous total internal reflection fluorescence spectroscopy and reflectance interference detection in a flow system as in principle described previously (25,26). Ternary complex formation was probed via the ligand-induced conformational change of IFNAR1 (27). For this purpose, IFNAR1-H10 N349C was cysteine-specifically labeled with ATTO 655 ( AT655 IFNAR1-H10), which is quenched by the proximal Trp 347 . Upon ligand binding, the accessibility of Trp 347 is substantially reduced, leading to dequenching of the dye (27). ATTO 655 was excited by a heli-um-neon laser (excitation power ϳ150 microwatts), and the fluorescence was detected through a bandpass filter. Continuous membranes were obtained by fusing lipid vesicles (steroyloleyl phosphatidylcholine), which were doped with ϳ5% lipids carrying tris-nitrilotriacetic acid head groups for tethering the receptor subunits (28) onto clean silica transducer slides. For comparing different combined IFN␣2 mutants, IFNAR2-H10 (ϳ3.5 fmol/mm 2 ) and AT655 IFNAR1-H10 (ϳ2.5 fmol/mm 2 ) were sequentially tethered onto the membrane. Subsequently, the ligand was injected (100 -500 nM), and its dissociation was monitored for 400 s, followed by an injection of imidazole (500 nM) for removing all proteins from the surface prior to the next experiment. The curves were drift-corrected by blank run subtraction, and the dissociation curves were fitted by a two-step dissociation model considering the two equilibria of the binary complexes IFN␣2-IFNAR1 and IFN␣2-IFNAR2 with the ternary IFN␣2-IFNAR1-IFNAR2 complex (26). The differential equations describing the two-step dissociation were fitted using Berkeley Madonna 8.0.1. The parameters for the interaction with IFNAR1 were kept constant for all IFN␣2 mutants, and only the two-dimensional dissociation constant for the interaction with IFNAR2 on the membrane was varied. As a control, the binary complex between AT655 IFNAR1-H10 and the IFN␣2 mutants was measured, and the dissociation curves were fitted by a monoexponential decay using BIAeval 3.1 software.
In Situ Binding Assays-Wild-type IFN␣2 was labeled with 125 I using the chloramine T iodination method (29). After iodination, the radiolabeled interferon was cleaned using a homemade Sephadex column. For the competition assay, WISH cells were grown on 24-well plates, washed once with phosphatebuffered saline plus 0.1% sodium azide, and then incubated for 10 min with the same solution. Next, cells were incubated for 1 h at room temperature (20°C) with the labeled wild type IFN␣2 (200,000 cpm/well) in the presence of an unlabeled IFN␣2 of interest at 10 different concentrations: wild type IFN␣2 and the YNS/ 153, YNS/148A mutants starting from 50 nM, YNS/30A and YNS/33A starting from 1 M, and the 30A, 33A, NLYY mutants starting from 6 M. All tested interferons were diluted in 6-fold steps in culture medium plus 0.1% sodium azide. Cells were then washed twice in phosphate-buffered saline to eliminate unbound interferons, trypsinated, and transferred into test tubes for measuring of bound, 125 I-labeled IFN␣2 wild type, using a ␥-counter (Packard). The experiment was repeated three times, each time in duplicate. IC 50 values were calculated using Kaleidagraph Synergy Software.
Antiviral and Antiproliferative Assays-The antiproliferative assay (16) was performed on WISH cells by adding the tested IFN␣2 (wild type or mutants) at serial dilutions to the growth medium in flat bottomed microtiter plates and monitoring cell density after 72 h by staining with crystal violet. The 50% activity concentrations (EC 50 ) as well as the sensitivity of cells to increasing amounts of interferon were deduced from an IFN dose-response curve (Kaleidagraph, Synergy Software) using Equation 2, where y represents the absorbance and reflects the relative Ternary Complex Stability Dictates Activity number of cells, A 0 is the offset, A is the amplitude, c is the IFN concentration, and s is the slope (16). Antiviral activity was assayed as the inhibition of the cytopathic effect of vesicular stomatitis virus on human WISH cells, as described previously (16,30). In general, IFN was added at serial dilutions to cells grown on flat bottomed 96-well plates. Four hours later, vesicular stomatitis virus was added to all wells, and after 17 h of incubation, cell density was measured by crystal violet staining. EC 50 was calculated as described for the antiproliferative experiment. Both the antiviral and antiproliferative assays were repeated at least three times for each protein. The experimental error () for both assays was 35%. Therefore, a confidence level of 2ϫ S.E. would suggest that differences smaller than 2-fold between interferons are within the experimental error.
pSTAT1 and pSTAT3 Phosphorylation Assays-WISH cells were lysed at 4°C with radioimmune precipitation buffer (150 mM NaCl, 10 mM Tris, pH 7.2, 0.1% SDS, 1% Triton X-100, 1% deoxycholate, 5 mM EDTA). Total protein concentrations were determined using a Bradford assay (Bio-Rad), and equal amounts were separated by SDS-PAGE. Levels of phosphorylated STAT1 and STAT3 were determined by Western blot analysis using polyclonal Tyr(P) 701 -STAT1-and Tyr(P) 705 -STAT3-specific antibodies (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Equal amounts of STAT1 and STAT3 protein were verified by reblotting with a polyclonal anti-STAT1 and -STAT3 antibody (Santa Cruz Biotechnology). Quantitative analyses of the Western blots were done using ImageJ software (National Institutes of Health).
Flow Cytometry Analysis of Cell Surface Receptors-Relative levels of IFN receptor subunits were assessed by indirect fluorescence immunostaining and fluorescence-activated cell sorting analysis, as described elsewhere (31,32). Surface IFNAR levels upon IFN treatment were quantified as relative to untreated cells (100%), using the median value of their signal histograms and taking the isotypic control levels as background.
Gene Expression Profiling-Total RNA was converted to cRNA, labeled, and hybridized on Agilent whole human genome 4X44K (G4112F) microarrays according to the manufacturer's protocol (Agilent). The slides were scanned in an Agilent DNA microarray scanner G2505B. Images were analyzed, and data were extracted using Agilent Feature Extraction software version 9.5.1.1, with linear and Lowess normalization performed for each array. A loop design with two biological replicates was designed to compare the four samples. The statistical analysis was performed using the Limma (Linear Models for Microarray Data) package from the Bioconductor project (available on the World Wide Web). The processed signal resulting from the Agilent Feature Extraction software was read into Limma using the "read.maimages" function. Aquantile normalization between arrays was applied. Standard quality control was performed using the plot functions of Limma (33). Differential expression was assessed using linear models for designed microarray experiments. The -fold changes and S.E. were estimated by fitting a linear model for each gene and applying empirical Bayes smoothing to the S.E. values (34). FDR (false discovery rate) was used to correct for multiple comparisons (35).

Design and Affinity Measurement of Combined IFNAR1/2
Binding Mutations-In all type I interferons, the binding affinity to the IFNAR2 receptor is much higher (up to 1000-fold) compared with the binding affinity to IFNAR1. This raises the possibility that the two receptors have differential roles in signaling, which is the tested hypothesis of the present study. We have recently constructed an optimized IFN␣2 mutant designated YNS, which binds IFNAR1 with ϳ60-fold higher affinity compared with wild type (K D ϭ 30 nM) while keeping its affinity toward IFNAR2 unchanged (K D ϭ 1.5 nM) (14). To examine the role of IFN binding to each receptor and correlate the differential affinities with differential biological outputs, we have constructed a series of Ala mutations on the IFNAR2 binding interface of IFN␣2 on the background of the YNS mutant, including Ser 153 , Met 148 , Leu 30 , and Arg 33 (Fig. 1A). These mutations were previously found to reduce the binding of IFN␣2 WT toward IFNAR2 by 6-, 60-, 700-, and 11,000-fold, respectively (36). Table 1 shows the binding affinities of these mutants as determined by SPR measurements, both on the YNS and WT background, toward IFNAR1 and IFNAR2. In accordance with the notion of independent IFNAR1/2 binding sites on IFN, the alanine mutations located on the IFNAR2 binding site of IFN␣2 reduced the affinity toward IFNAR2 to the same extent both in the WT and YNS backgrounds. Accordingly, the affinity toward IFNAR1 of all YNS-containing mutant proteins increased by 50 -60-fold, regardless of the reduction in IFNAR2 binding. Overall, these mutants provide proteins with IFNAR2/IFNAR1 binding affinity ratios from 1000:1 down to 1:1000 toward the two receptors, with the YNS/153A mutant having equal binding affinities to both receptors.
Ternary Complex Stability as Measured on Artificial Membranes-In order to explore the consequences of the changes in ligand binding affinities toward IFNAR1 and IFNAR2, we probed ternary complex formation on artificial membranes by simultaneous total internal reflection fluorescence spectroscopy-reflectance interference detection. IFNAR1-EC and IFNAR2-EC were tethered to the membrane via their C-terminal His 10 tags, which were selectively captured by tris-nitrilotriacetic acid moieties embedded into the membrane, enabling lateral diffusion and ligand-induced ternary complex formation (23,26,28). IFNAR1-EC N349C cysteinespecifically labeled with ATTO 655 ( AT655 IFNAR1-EC) was used for probing binding. The ligand-induced conformational change of IFNAR1-EC strongly reduces the accessibility of Trp 347 , leading to dequenching of the dye (27). A typical assay is shown in Fig. 1B; after sequential tethering of IFNAR2-EC and IFNAR1-EC, the ligand is injected and detected by the increase in fluorescence of AT655 IFNAR1-EC. Whereas on the reflectance interference channel, all ligand binding to the receptor subunits on the membrane is detected, the fluorescence channel selectively detects binding to IFNAR1.
The dissociation curves for different IFN␣2 mutants from the ternary complex are compared in Fig. 1C. With decreasing affinity toward IFNAR2, faster dissociation from IFNAR1 in the presence of IFNAR2 is observed. On the contrary, in the absence of IFNAR2 on the surface, very similar dissociation from IFNAR1-EC was confirmed for all IFN␣2-YNS mutants (Fig. S1A). Substantial stabilization of ligand binding by the presence of IFNAR2 was observed even for IFN␣2-YNS/33A (blue versus gray in Fig. 1C). Although the quenching of AT655 IFNAR1-EC is not directly related to the half-life of the ternary complex (since it is a measure of IFN-IFNAR1 binding), it is qualitatively related. Thus, the half-life extracted from the dissociation curves of the different mutants (Table 2) provides a measure of the stability of the ternary complex. The important role of receptor surface concentration is demonstrated in Fig.  1D for IFN␣2-YNS/33A; a 3-fold increase in the half-life of the ternary complex is observed upon a 4-fold increase of the receptor surface concentration. Comparison of the dissociation kinetics of IFN␣2 WT from the ternary complex measured at a similar receptor surface concentration yields a curve between the curves obtained for IFN␣2-YNS/153A and IFN␣2-YNS/ 148A (Fig. S1B). A schematic representation of the dynamic stabilization of the ternary complex is presented in Fig. 1E.
In Situ Measurements of Ternary Complex Stability on the Cell Surface-Formation of a ternary IFN-IFNAR1/2 complex on the cell surface is necessary to relay an IFN signal into the cell. SPR was used to measure in vitro the binding affinities of the different mutants toward each receptor individually, and the measurements on artificial membranes provide us with in vitro data on the ternary complex stability. The relevance of the in vitro data was then evaluated with in situ measurements of the various interferon mutants toward the IFNAR receptors. IC 50 values were determined using a binding competition assay on the cells with 125 I-labeled IFN␣2 wild type mixed with cold IFN (either wild type or mutants), measuring residual 125 I-labeled IFN␣2 binding. The IC 50 of binding of the four combined mutants as well as L30A, R33A, and NLYY (Ala mutations in residues 65, 80, 85, and 89 resulting in almost no binding to IFNAR1) resulted in micromolar to picomolar values (Table 2 and Fig. S2). Relating the in situ EC 50 values to the combined affinities of IFNs to both receptors (R1 ϫ R2; Table 2 and Fig.  2A) as well as to the two receptors bound on artificial membranes ( Fig. 1 and Table 2) showed a positive relation between the data. The logic of calculating a combined affinity (R1 ϫ R2) from the individual binding constants is that this represents the simple case where each receptor contributes additively toward the total binding affinity of both receptors to interferon, as seems to be the case on the cell surface. Moreover, the IC 50 of the wild-type interferon is similar to that measured for the 148/ YNS mutant, in agreement with the data presented in Tables 1  and 2 for the combined binding affinities or for binding to both receptors anchored on artificial membranes. This suggests a simple binding model of interferon to both receptors, in which the combined affinity is a function of the additive value of the individual contributions. This concept is further supported by the IC 50 value for the NLYY mutant (4 nM), which is close to the in vitro binding constant determined for IFNAR2 alone using SPR. The half-life of the NLYY mutant as measured in Fig. 1 on artificial membranes was 50 s (Table 2), which is very similar to the value measured for wild-type interferon to IFNAR2 alone, in line with the very weak binding of this mutant to IFNAR1.
The weakest IFNAR2 binding mutant tested was R33A, reduc-FIGURE 1. Ternary complex formation on artificial membranes. A, ternary structure of the IFN-IFNAR1-IFNAR2 complex (18). Marked in blue are mutations facing IFNAR2, and marked in red are mutations facing IFNAR1 (YNS). B, typical binding assay as detected on the reflectance interference channel (black) and the total internal reflection fluorescence spectroscopy channel (red). I, tethering of IFNAR2-H10; II, tethering of AT655 IFNAR1-H10; III, binding IFN␣2 YNS; IV, removal of the proteins by imidazole. C, ternary complex dissociation of YNS-IFN␣2 bearing mutations at the IFNAR2 binding site in comparison with the dissociation from IFNAR1-H10 alone. All experiments were carried out with ϳ2.5 fmol/mm 2 AT655 IFNAR1-H10 and ϳ3.5fmol/mm 2 IFNAR2-H10. Black lines correspond to the best fit by a two-step dissociation model. For comparison, dissociation of YNS/33A from AT655 IFNAR1-H10 alone measured under the same conditions is shown in gray. D, dissociation of IFN␣2 YNS/33A at two different receptor surface concentrations of AT655 IFNAR1-H10 and IFNAR2-H10: ϳ12.5 fmol/mm 2 and ϳ20 fmol (red) and ϳ2.5 fmol/mm 2 and ϳ3.5 fmol (blue). For comparison, dissociation of IFN␣2 YNS/33A from IFNAR1-H10 alone is shown (black). E, schematic drawing of the IFN-induced assembly of the ternary complex, leading to a dynamic equilibrium between the binary IFN-IFNAR1 and IFN-IFNAR2 and the ternary IFN-IFNAR1-IFNAR2 complexes. The dynamics of the exchange of subunits is determined by the two-dimensional rate constants k d, 1 T and k d,2 T (25). The total binding affinity is determined by binary complex equilibrium constants K D,1 B and K D,2 B as well as the equilibrium between binary and ternary complex, which is determined by K D, 1 T , K D,1 T and the receptor surface concentrations. Altering the binding affinities toward the receptor subunits will affect not only the total binding affinity toward the ligand but also the equilibria between ternary and binary complexes and the dynamics of the exchange.

TABLE 1 In vitro binding affinities of interferon wild type and mutants towards IFNAR1 and IFNAR2 receptor subunits
Binding was measured using ProteOn XPR36. K D values were determined from k d /k a over six different concentrations of the analyte. Ratios are relative to wild type IFN␣2 (given in the first row). The affinity for IFN␣2 was determined using the mass-action equation over six different concentrations of the analyte. The S.E. value for k a , k d , and K D values are 12, 25, and 35%, respectively. ND, below the detection limit.

IFNAR1
IFNAR2  Antiproliferative and Antiviral Activities Are Differently Affected by Mutations on Interferon-The biological activity of the different mutants was tested in WISH cells. In this cell line, the initiation of the antiviral state requires at least 4 h of IFN treatment at pM concentrations, whereas antiproliferative (apoptotic) activity is measured only after 72 h and requires IFN in nanomolar concentrations. In this cell line, the YNS mutant is a much stronger activator of the antiproliferative effect than wild-type IFN␣2 (100-fold increase), whereas its EC 50 value for antiviral activity is hardly changed (Table 3). We were therefore interested in determining whether IFNAR1 has an exclusive effect on the antiproliferative activity. Antiproliferative and antiviral potencies of the interferon mutants in WISH cells were determined from the EC 50 values as calculated from dose-response curves ( Table 3). The antiproliferative activity correlated with the in situ binding affinity (Fig. 2B), whereas for the antiviral activity, this correlation was observed only for affinities lower than wild type. In addition, different slopes of 0.53 and 0.88 were observed for the antiviral and antiproliferative effects, respectively. Thus, the antiviral activity seems to approach a saturated phase, where it is less sensitive to further changes in the affinity, as opposed to the antiproliferative activity that is highly affected by affinity changes throughout the entire given range. In addition, the NLYY mutant, which has no measurable IFNAR1 binding in vitro (16), exhibited a distinctive behavior with activities lower than one would expect from its overall in situ affinity. These data are in line with our notion that NLYY binds normally to IFNAR2 (with nanomolar affinity) but binds IFNAR1 very weakly and therefore hardly promotes the formation of the ternary complex, which is critically required for biological activity.
STAT1 Phosphorylation Follows the Same Pattern as Antiviral, but Not Antiproliferative Activity-We next sought to define the relation between the binding affinities of the IFN mutants and phosphorylation of STAT1, a major signaling factor in the JAK-STAT pathway associated with IFN signaling. pSTAT1 levels on tyrosine 701 reach a maximum after the first 1 h of interferon induction, after which pSTAT1 declines back to noninduced levels (14,32). On the contrary, the total STAT1 (and STAT2 but not STAT3) protein levels rise only after more than 8 h of continuous induction (32). Therefore, WISH cells were treated with the various mutants at different concentrations (2.5-800 pM) for 45 min in order to assess pSTAT1 by Western blot using Tyr(P) 701 -STAT1specific antibodies (Fig. 3A). For all measurements, levels of general protein were similarly assessed using the appropriate antibodies, showing no increase in protein concentrations for all treatments. As seen in Fig. 3A, STAT1 phosphorylation levels were similar for YNS, WT, YNS/153A, and YNS/ 148A and reduced for the other mutants, with no pSTAT1 observed at concentrations up to 800 pM for the R33A mutant. The same pattern of induction was observed also for STAT3 phosphorylation (Fig. 3A, right). To further quantify the levels of pSTAT1, the Western blots shown in Fig. 3A were scanned, quantified, and normalized against total protein levels (Fig. 3B). Phosphorylation values were then plotted versus the IFN concentration, from which the EC 50 values were determined using Equation 2 and are shown in Table 3. No large differences in EC 50 for pSTAT1 or pSTAT3 relative to WT were detected for YNS and YNS 153A (Table  3 and Fig. 3A). Comparing the phosphorylation levels of STAT1 and STAT3 with the activity profiles for each mutant (Table 3 and Fig. S3) clearly shows that the antiviral potency FIGURE 2. The effect of interferon binding on its activity. A, theoretical versus measured affinities of different interferon mutants toward the receptor complex. X axis, binding affinities as measured in situ in WISH cells; y axis, calculated affinities based on the contribution of each receptor subunit alone (R1 ϫ R2), as measured in vitro using SPR. B, relative biological potency versus relative in situ affinity of IFN␣2 mutants. X axis, in situ binding affinity relative to wild type IFN␣2. Y axis, antiproliferative (AP) (f) and antiviral (AV) (Ⅺ) potency of combined IFN␣2 mutants relative to wild type. Crosses represent an IFN mutation with almost no IFNAR1 binding but normal IFNAR2 binding.

TABLE 3 Biological activities and STAT1 phosphorylation
Ratios are relative to wild type interferon (given in the first row). of these mutants follows a similar trend as pSTAT1 and pSTAT3, whereas the antiproliferative activity does not. Moreover, both pSTAT1 levels and antiviral potency reach a maximum once the EC 50 of interferon binding to the surface receptor reaches the level of wild type IFN␣2 and do not increase further (Table 3 and Fig. S3). On the contrary, the antiproliferative activity of the different mutants is not related to the pSTAT1 activation pattern.

Differential Down-regulation of the IFNAR Receptor Subunits Is Dictated by the Integral Lifetime of the Receptor-IFN Complex-
IFNAR receptor down-regulation is generally accepted as a means of signal attenuation following an IFN stimulus. In a recent report, it was shown that cell surface IFNAR internalization is actually required for a full and fast transduction of the IFN signal (37). Treatment with different IFN subtypes, namely IFN␣2 versus IFN␤, resulted in a differential down-regulation of the IFNAR subunits (32). Particularly, IFNAR1 was down-regulated after 16 h of treatment with IFN␤ but not IFN␣2. This finding raised a question regarding the role of internalization in differential IFN signaling. Receptor numbers on the surface of WISH cells were quantified here by fluorescence-activated cell sorting following interferon treatment. We first compared IFN␣2 with IFN␤ at 30 pM, a concentration below a full antiproliferative effect in WISH cells for both IFNs, and 300 pM, at which only IFN␤ can promote a full antiproliferative response (31). During the course of IFN treatment at low concentrations, both IFNAR subunit levels were initially down-regulated, followed by a linear recovery over time, as expected (Fig. 4A).   NOVEMBER 21, 2008 • VOLUME 283 • NUMBER 47

Ternary Complex Stability Dictates Activity
However, at 300 pM IFN␤ (but not IFN␣2), the down-regulation of IFNAR1 was sustained for the entire length of the experiment (24 h), whereas surface expression levels of IFNAR2 returned to normal. To eliminate reduced IFNAR transcript levels as a possible explanation for the observed down-regulation, we analyzed receptor mRNA levels on an Agilent gene chip (see Fig. 5) and found no change in IFNAR1 and a small (1.6-fold) increase in IFNAR2 mRNA levels after 36 h of treatment, independent of the interferon used.
Using the combined mutants, we were able to test whether the sustained IFNAR1 down-regulation was the result of a higher affinity of IFN toward IFNAR1 or a higher affinity toward the ternary complex in general. If high affinity toward IFNAR1 alone is sufficient, these combined mutants will still induce prolonged down-regulation of IFNAR1, similar to YNS or IFN␤. We measured the levels of cell surface IFNAR subunits upon incubation with these IFN mutants at 100 pM, a concentration that is sufficient to induce a maximal antiproliferative response with IFN␤ and YNS but not with IFN␣2. Again, the early down-regulation of IFNAR2 was followed by linear recovery (Fig.  4B). Only IFN␤ and the YNS mutant induced a continuous differential IFNAR1 down-regulation at this concentration. Strikingly, the YNS/ 148A mutant and IFN-␣2 WT, which have comparable combined affinities, induced similar IFNAR1 down-regulation patterns. The YNS/30A mutant promoted the weakest IFNAR1 down-regulation, in agreement with its weak combined affinity. These results show that a high affinity toward IFNAR1 is not sufficient to induce its prolonged down-regulation. Rather, differential receptor down-regulation is dictated by the integral lifetime of the ternary receptor-IFN complex. This can result from high binding affinity to the receptors, such as in the case of YNS or IFN␤, or high concentration of weaker binding interferons, such as 3 nM IFN␣2, which also induced prolonged IFNAR1 down-regulation as well as antiproliferative activity (data not shown). Importantly, these data exhibit an intriguing correlation between differential receptor down-regulation and the antiproliferative potency.
IFN-induced Gene Regulation-We have previously shown that after 16 h of treatment with interferon, the gene induction profile of the IFN␣2 triple mutant HEQ (57A, 58A, 61A) is similar to that of IFN␤ and quantitatively, but not qualitatively, different from that observed for IFN␣2 (31). Bearing in mind that for a cytostatic effect, a prolonged exposure of target cells to IFN is required, we analyzed here the gene induction profile after 36 h of induction with 100 pM IFN␣2, YNS, or IFN␤. This concentration was chosen, since it promotes antiproliferative activity using YNS and IFN␤ but not using IFN␣2 (yet is far above the concentration required for a full antiviral activity for all three) ( Table 3) (14). Contrary to our observations following 16 h of IFN treatment, at 36 h we already observed a qualitative difference in gene activation. Of particular interest are those genes that are differentially up-regulated at 36 h of YNS or IFN␤ but not IFN␣2 treatment (Fig. 5A). By sorting the expression data according to their -fold change upon YNS or IFN␣2 induction, we found a set of 135 genes that were differentially up-regulated (at least 3-fold stronger for YNS compared with IFN␣2) and 188 genes that were equally up-regulated (by at least 3-fold) in both treatments (Fig. 5A). induced genes (open squares) and a set of 188 equally induced genes (solid circles) were retrieved. A, normalized gene induction by YNS treatment (y axis) plotted against normalized gene induction by IFN␣2 WT (x axis, left plot) or IFN␤ (x axis, right plot). For both sets of genes, no significant change in the level of induction is observed between YNS and IFN␤. B, the two sets of genes were subjected to gene annotation analysis using the bioinformatics resource DAVID (available on the World Wide Web), which identified enriched biological processes. These were clustered using the cluster algorithm in DAVID and marked by colors according to cluster.
Next, we used the DAVID bioinformatics resource (available on the World Wide Web) to functionally annotate these two gene sets. Fig. 5B shows the ϳ20 most up-regulated gene functions, clustered into six families (sorted by DAVID). The two gene sets are distinct in their functions. The equally up-regulated genes code for genes involved in direct response to viruses and other pathogens as well as for up-regulation of the major histocompatibility complex I response. Conversely, the YNS differentially up-regulated genes code for inflammation and cytokine and chemokine functions. Both sets have in common the activation of the immune system. Although the difference in gene activation is quantitative, the prolonged YNS versus IFN␣2 treatments code for a qualitative difference in the way the cellular defense is activated.

DISCUSSION
The differential activation of cells by IFNs through a shared cell surface receptor calls for a molecular explanation. IFNs exert their activity through the binding to two different receptor subunits, and it has been suggested that each subunit possesses specific signaling capabilities. Recent findings have established a key role for ligand binding affinity toward IFNAR1 in differential activation (11,31,37,38). Here, we have aimed to systematically define the relation between ligand binding affinities toward the individual subunits, the formation of the ternary complex, and the activation of signaling pathways and different cellular responses. For this purpose, we engineered a panel of IFN␣2-based mutants binding IFNAR1 and IFNAR2 at affinity ratios ranging from 1000:1 to 1:1000. These mutants cause a reduction in IFNAR2 binding to different extents, whereas IFNAR1 binding is significantly enhanced using the IFN␣2-YNS triple mutant that binds IFNAR1 60-fold tighter than wild type. We also used the NLYY mutant, where IFNAR1 binding is reduced by ϳ100-fold. Affinity changes toward the two receptor subunits were verified by SPR using purified receptor subunits, on solid-supported membranes with the two receptors immobilized, and on the cell surface using 125 I-labeled interferon. The affinities of these ligands toward both receptors in vitro and on the cell surface correlated well with the theoretically estimated affinity from the contribution of each receptor subunit, corroborating independent ligand binding sites on IFNAR1 and IFNAR2. This is in line with the distinct location of the binding sites of the two receptor subunits on interferon (Fig. 1A).
Using these mutants, we observed a clear linear correlation between the relative antiproliferative potencies and the relative cell surface receptor binding affinities with a slope of ϳ1. These findings suggest a simple additivity, rather than cooperativity, in the binding of the two receptors and establish that the total binding affinity, but not the particular affinity toward one of the receptor subunits, dictates the antiproliferative activity of IFNs. A rather simple model for antiproliferative IFN signaling emerges from these results, which is determined by two factors: the rate of complexation events and the stability of the ternary complex once it is formed. The latter is a function of the affinity toward both receptors and the concentration of all components (Fig. 6). Thus, high interferon or receptor concentrations may compensate for low binding affinities, as is indeed observed (11,31). In contrast, the relationship between the relative antiviral potencies of the different mutants and the binding affinities is more complex. Antiviral activity reaches a level close to maximum (in WISH cells) when the combined affinity to both receptors (R1 ϫ R2) is that of wild-type IFN␣2 (14). A similar phenomenon of maximal cellular activity independent of the affinity was previously found for human growth hormone (39). However, we show here that the maximal antiviral activity is related to the overall receptor binding affinity, independent of whether it results from a tighter IFNAR1 or IFNAR2 binding. This is clearly demonstrated by the YNS/153A and YNS/148A mutants; these have a wild-type-like antiviral potency, dictated by their total binding affinity to the cell surface receptor that is close to the wild-type protein, although IFNAR1 binding is strengthened and IFNAR2 binding is weakened in these mutants. Interestingly, the slope of the antiviral-affinity relation is only ϳ0.6 and not close to the expected slope of 1, as observed for the antiproliferative activity.
Since these differences in the affinity-activity correlations for antiviral and antiproliferative responses may define the differential signaling patterns of IFNs, we analyzed the downstream consequences of differential receptor binding affinities. We found that pSTAT1 or pSTAT3 activation relates to the stability of the ternary complex, with a similar trend as the antiviral but not antiproliferative activity (Fig. S3). These data support the notion that accumulation of pSTAT1 and pSTAT3 is directly related to the initiation of an antiviral state (Fig. 6). In contrast, antiproliferative activity seems to be related not only to the STAT pathway but to an additional STAT-independent pathway, which is activated in a linear activity-affinity relationship.
An interesting characteristic of the antiviral activity lies in its low EC 50 value, which is 0.3 pM for IFN␣2, whereas the EC 50 of in situ receptor binding and pSTAT1 activation is ϳ50 pM. These results suggest that activation of a minor part of the cell surface receptors and thus the induction of submaximal STAT phosphorylation is sufficient to promote full antiviral protection (ϳ1% receptor occupancy provides 50% protection). In contrast, antiproliferative activity requires ligand concentrations well above the EC 50 of the cell surface receptor binding. We suggest that these observations can be explained by the kinetics of receptor binding and dissociation. We have previously demonstrated on artificial membranes that the ternary IFN-receptor complex is in dynamic equilibrium with the binary complexes of IFN with IFNAR1 and IFNAR2, as schematically depicted in Fig. 1E (25). Differences between the binding affinity of IFNs toward the receptor subunits will change the equilibrium concentrations of binary and ternary complexes as well as the dynamics of the equilibrium. The lifetime of the ternary complex is limited by the shorter lifetime of the interaction with either one of the receptor subunits, leading to a faster exchange of the receptor subunits. It appears that STAT phosphorylation (and thus the induction of an antiviral response) does not require a stable ternary complex; even for YNS/33A, with an estimated lifetime of ϳ20 ms for the IFN-IFNAR2 interaction, some STAT phosphorylation is observed. Since phosphorylation is a very fast process (millisecond range), longer complex lifetimes are probably not required. It is even possible that rapid exchange of receptor subunits promotes STAT phosphorylation, since mutants with a short lifetime of interaction with IFNAR2 (e.g. IFN␣2 L30A) show relatively high STAT phosphorylation and antiviral activity compared with their binding affinity to the cell surface receptor.
To further quantify the STAT1 phosphorylation pattern, we simulated the rate of accumulation of pSTAT1 in cells using the Pro-Kineticist II software, a second order global kinetic analysis software (Applied Photophysics Ltd.). We used the model presented in supplemental Scheme 1, which contains in addition to the equilibrium between the different forms of the interferonreceptor complex (as presented in Fig. 1E) also a term for activation (by Tyk2) and deactivation (by either SOCS or receptor endocytosis) of the phosphorylated, active ternary complex. The simulations show similar pSTAT1 50 levels (the IFN concentration giving 50% pSTAT1) for wild-type IFN␣2, YNS, YNS/153A, and YNS/148A. Weaker binding mutants, such as YNS/30A, L30A, and NLYY, have lower calculated pSTAT1 50 values, which correlate well with the experimental data. The predictive nature of the calculations was independent of whether the affinity to IFNAR1, IFNAR2, or both were altered. The simulations suggest that the rate of deactivation of the active (STAT1 binding) ternary complex becomes the rate-limiting factor once the lifetime of the ternary complex exceeds that observed for wild-type IFN␣2. Therefore, further stabilization of this complex by mutations, such as YNS (as seen in Fig.  1), has no effect on STAT1 phosphorylation. However, for weaker binding mutants, the lifetime of the active receptor complex is dictating the rate of pSTAT1 accumulation. The different rate-limiting steps along the reaction are also the reason for the slope of less than 1 observed for the relation between binding affinity and pSTAT1 accumulation (and antiviral activity).
Rapid, ligand-induced down-regulation of both IFNAR1 and IFNAR2 (endocytosis) has been previously reported (40 -43). This down-regulation was indeed shown to affect IFN signaling, both by signal attenuation (38) and direct involvement in signal propagation (37). Reduction in cell surface concentrations of one or both of the receptor subunits reduces cell sensitivity and alters further signaling (37,38,40,44,45). Moreover, a clear correlation was suggested between the efficiency of interferon as a cancer drug and surface receptor concentration (38,46,47). In a recent study, Marijanovic et al. (40) showed that the cell surface IFNAR1 level is indeed a limiting factor for assembly of the functional complex, but an increased concentration of it does not translate into an IFN␣/␤ differential JAK-STAT signaling; nor does it change the dynamics of the engaged receptor. Tyk2 and receptor ubiquitination were identified as the main factors controlling the endocytosis of IFNAR and its routing to degradation versus recycling (48 -50). However, all of these studies probed only the first few hours after interferon induction and not the long term effects described in this paper. Rapid receptor down-regulation and STAT activation is observed on a time scale of hours. However, antiproliferative activity works on the time scale of days (Fig. 6). We found here that IFNAR1 and IFNAR2 follow a differential down-regulation pattern after continuous treatment for 36 h. Upon treatment with either IFN␣2 or IFN␤, membrane, IFNAR2 is rapidly down-regulated but returns to normal concentrations within 36 h of treatment. On the contrary, IFNAR1 down-regulation is more complex. Following 2 h of treatment, IFNAR1 is down-regulated, independent of the interferon used. However, although IFNAR1 cell surface concentrations recover for IFN␣2-treated cells with a similar kinetics as observed for IFNAR2, IFN␤or YNS-treated cells maintain substantially lower concentrations of IFNAR1 (around 40% of basal level) even after 36 h of treatment. In light of the high affinity of IFN␤ and YNS toward IFNAR1, we initially suspected that IFNAR1 binding by itself might account for this phenomenon. However, fluorescence-activated cell sorting measurements of the combined mutants (YNS/148A or YNS/30A) revealed that IFNAR1 down-regulation correlates with the overall stability of the ternary complex (which is limited by the weakest interaction) and not necessarily with the stability of IFN-IFNAR1 interaction. We therefore conclude that the long term differential pattern of receptor down-regulation is proportional to both receptor subunit affinities in an additive manner, thus rejecting the possibility of a unique role for IFNAR1 in initiating this pattern. These low expression levels of IFNAR1 on the membrane are likely to reduce responsiveness of the cells to some IFNs, perhaps giving an advantage to strong IFNAR1/2 binders. A similar down-regulation pattern was shown in human dendritic cells by Severa et al. (44), where autocrine release of IFN␤ caused a long term reduction of IFNAR1 but not IFNAR2 levels on the cell membrane. These cells were indeed desensitized to IFN␣2 treatment, whereas their responsiveness to IFN␤ treatments remained normal. Importantly, the differential IFNAR1 down-regulation pattern closely follows the antiproliferative signal in terms of signal duration, affinity toward the receptor, and interferon concentrations needed for prolonged IFNAR1 down-regulation, whereas pSTAT1 activation follows the antiviral signal. Along with our observation that the antiviral potency of IFNs, but not their antiproliferative potency, is naturally close to saturation levels, these observations support the idea of a pSTAT1-independent pathway(s) required for antiproliferative signaling (Fig. 6). Nuclear localization and transactivation activity was recently reported for IFN-␥ and its IFNGR1 receptor subunit (51), and since IFNAR1 contains a nuclear leading sequence (41), it is likewise tempting to speculate that following endocytosis, this receptor subunit acts as an intracellular signaling factor, mediating the expression of key antiproliferative genes. Further investigation into the fate of IFNAR1 once internalized in the cell may provide important insights on the differential signaling of IFNs.
Gene activation patterns after 36 h of constant treatment with IFN confirmed a clear qualitative difference between YNS-and IFN␣2-induced gene activation, with YNS inducing a similar pattern as IFN␤. Those genes that are up-regulated by all three IFNs relate to antiviral and major histocompatibility complex I functions, whereas the differentially up-regulated genes (by YNS and IFN␤) relate to inflammation and cytokine and chemokine responses. These results strengthen our notion that differential interferon action is indeed related to the binding affinity toward the ternary complex, the time of activation, and the concentration of the interferon and that these dictate a clear difference in the physiological outcome.
In summary, systematic variation of subunit binding affinity has revealed that the nonlinear correlation of antiviral activity and cell surface receptor binding leads to differential activity patterns of IFN (Fig. 6). We speculate that the stabilization of the ternary complex on the membrane, together with the rate of deactivation of the signaling complex lead to the nonlinearity of pSTAT1 activation, which in turn leads to the nonlinear effect (in respect to binding affinity) of the antiviral response. Moreover, we postulate a pSTAT-independent pathway responsible for antiproliferative activity, which strictly depends on ligand binding affinity and is correlated with IFNAR1 down-regulation. A differential set of genes is activated by this pathway, yet so far we could not identify a common transcription factor driving differential gene production. However, whether binding is tighter to IFNAR1 or IFNAR2 is not a major factor in differential activation.