View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector

Structure, Vol. 11, 791–802, July, 2003, 2003 Elsevier Science Ltd. All rights reserved. DOI 10.1016/S0969-2126(03)00120-5 The Human Type I Receptor: NMR Structure Reveals the Molecular Basis of Ligand Binding

Jordan H. Chill,1 Sabine R. Quadt,1 Rina Levy,1 mia (CML) (Bukowski et al., 2002), and hepatitis C (Perry Gideon Schreiber,2 and Jacob Anglister1,* and Jarvis, 2001). Other ␣-IFNs have been approved 1Department of Structural Biology for treatment of additional hepatitis subtypes, chronic 2 Department of Biological Chemistry malignant melanoma, follicular lymphoma, and genital Weizmann Institute of Science warts caused by the human papilloma virus. IFN␤ is Rehovot 76100 FDA approved for the treatment of relapsing-remitting Israel multiple sclerosis (Deonarin et al., 2002). However, effi- cient treatment by IFNs is frequently hampered by vari- ous side effects (e.g., toxicity), neutralizing antibodies Summary and cell resistance (Einhorn and Grander, 1996). The efficacy of IFN treatment could be enhanced if these The potent antiviral and antiproliferative activities of undesirable effects could be prevented or reduced with human type I (IFNs) are mediated by a sin- a more selective and optimized activity profile. gle receptor comprising two subunits, IFNAR1 and Type I IFNs share a common receptor consisting of IFNAR2. The structure of the IFNAR2 IFN binding ecto- two subunits, IFNAR1 (Uze et al., 1990) and IFNAR2 domain (IFNAR2-EC), the first helical (Novick et al., 1994), which associate upon IFN binding structure determined in solution, reveals the molecular (Cohen et al., 1995). IFNAR2 is the major ligand binding basis for IFN binding. The atypical perpendicular orien- component of the receptor complex, exhibiting nanomo- tation of its two fibronectin domains explains the lack lar affinity to both IFN␣ and IFN␤ subtypes. This affinity of C domain involvement in ligand binding. A model is increased up to 20-fold upon formation of the ternary of the IFNAR2-EC/IFN␣2 complex based on double complex with IFNAR1 (Cohen et al., 1995). The extracel- mutant cycle-derived constraints uncovers an exten- lular domain of IFNAR2 (IFNAR2-EC) was expressed in sive and predominantly aliphatic hydrophobic patch E. coli. The purified and folded IFNAR2-EC was found on the receptor that interacts with a matching hy- to be a stable , which retains the full binding drophobic surface of IFN␣2. An adjacent motif of alter- activity of IFNAR2 and inhibits the antiviral activity of nating charged side chains guides the two IFN␣2 by competing with the cellular IFN receptor for into a tight complex. The binding interface may ac- IFN binding (Piehler and Schreiber, 1999a). count for crossreactivity and ligand specificity of the The structures of IFNAR1 and IFNAR2, both class II receptor. This molecular description of IFN binding helical cytokine receptors (HCRs), have not been solved should be invaluable for study and design of IFN-based to date. The backbone resonances of IFNAR2-EC were biomedical agents. assigned, and residues involved in IFN binding were identified, by following ligand binding-induced changes Introduction in IFNAR2-EC chemical shifts (Chill et al., 2002). This investigation, as well as mutagenesis (Lewerenz et al., Type I interferons (IFNs) are a family of homologous 1998; Piehler and Schreiber, 1999b) and immunoblock- helical cytokines that elicit potent antiviral and antiprolif- ing (Chuntharapai et al., 1999) studies, located the IFN erative cellular responses. IFN binding to its receptor binding site on three segments of the N-terminal domain triggers a cascade of events, activating a number of of IFNAR2-EC, including residues 44–52, 74–82, and proteins that inhibit viral replication and cell growth and 100–104. Residues of IFN ligands contributing to the control apoptosis (Stark et al., 1998). IFNs are essential binding energy were identified as well (Piehler et al., for the survival of higher vertebrates because they pro- 2000; Runkel et al., 2000). Models for IFNAR2-EC based vide an early line of defense against viral infection, hours on its limited homology with tissue factor (TF; 21% iden- to days before the immune response (Stark et al., 1998). tity) (Harlos et al., 1994), as well as other helical cytokine This family of cytokines consists of about fifteen differ- receptors, such as the class II IFN␥ receptor (IFN␥R; ent IFN␣ isotypes, typically exhibiting 80% sequence 17% identity) (Walter et al., 1995) and the class I growth homology, IFN␤, exhibiting 30% identity to a consensus (GHR; 14% identity) (de Vos et al., IFN␣ sequence, and IFN␻ (Pestka et al., 1987). The struc- 1992), have been proposed (Chill et al., 2002; Piehler tures of IFN␣2a (Klaus et al., 1997), IFN␣2b (Radhakrish- and Schreiber, 1999b). One such model was used to nan et al., 1996), and IFN␤ (Karpusas et al., 1997) have dock the IFN␣2 ligand into its binding site on IFNAR2- been previously determined, showing IFNs to comprise EC with double mutant cycle data (Roisman et al., 2001). a bundle of five long and nearly parallel ␣ helices. However, the accuracy of the IFNAR2-EC models was IFNs are currently the human proteins most widely severely limited by the low between used as therapeutic agents, having been approved to receptors of this family and several insertions and dele- treat various types of cancer and viral diseases. IFN␣2is tions in their sequences (seven insertions and deletions used to treat hairy cell leukemia, AIDS-related Kaposi’s involving 32 residues for IFN␥R, the most similar recep- sarcoma (Kirkwood, 2002), chronic myelogenous leuke- tor). The model of IFNAR2-EC was improved with NMR- derived secondary structure information (Chill et al., *Correspondence: [email protected] 2002). Nevertheless, the difficulties in modeling IFNAR2- Structure 792

EC have frustrated attempts to understand binding of Table 1. Structural Statistics for IFNAR2-ECa IFN ligands by their receptor on a molecular level. One of the major issues in IFN research is how the NMR Distance Restraints versatile IFNAR2 specifically binds different IFN sub- Total interresidue restraints 1876 types, thereby transducing different cellular responses Sequential (|i Ϫ j| ϭ 1) 561 (Mogensen et al., 1999). The relative contribution of both Medium range (1 Ͻ |i Ϫ j| Ͻ 5) 280 | Ϫ | Ն receptor subunits to ligand crossreactivity and signaling Long range ( i j 5) 1035 Hydrogen bond restraintsb 134 specificity is currently unclear. Previous studies have NOE violations (A˚ ) suggested partially overlapping binding sites upon Maximum single violation 0.5 IFNAR2 for the IFN␣ and IFN␤ subtypes and different Rmsd of NOE violationc 0.055 Ϯ 0.001 contributions of specific residues to the exhibited affin- Dihedral Angle Restraints ity. Mutations of human IFNAR2 residues RE77, RW100, RI103, and RD104 abolished response to IFN␣2 with no Total dihedral restraints 180 Experimental ( )88 effect upon IFN␤ binding (Chuntharapai et al., 1999). φ TALOS derived ( ,␺)92 ␣ ␤ φ (Residues of IFNAR2-EC, IFN 2, and IFN are preceded Dihedral angle violations (Њ) by superscripts R, ␣, and ␤, respectively, throughout Maximum single violation 5 the text.) Schreiber and coworkers found that, while Rmsd of dihedral violation 0.760 Ϯ 0.036 R R R ␣ T44, I45, and M46 are the key residues for IFN 2 Deviation from Idealized Geometryd binding (with some additional contribution from RS47, R R R R R R R Bond lengths (A˚ ) 0.004 Ϯ 0.000 K48, H76, E77, W100, and I103), I45 and W100 Њ Ϯ ␤ Bond angles ( ) 0.600 0.010 are the major contributors to IFN binding (with smaller Improper angles (Њ) 0.560 Ϯ 0.016 contribution by residues RT44, RM46, RS47, and RI103) ˚ (Piehler and Schreiber, 1999b). Furthermore, important Mean Rmsd Values (A) differences between IFN subtypes correlated to signal- Backbone 0.67 ing profiles have been identified on the opposite face of Heavy atoms 1.12 the ligand, a surface postulated to interact with IFNAR1 Backbone (2Њ structure) 0.50 Њ (Deonarin et al., 2002; Runkel et al., 2000). Heavy atoms (2 structure) 0.90 To examine IFN binding and signaling at the molecular a Excluding N-terminal residues 1–11 and C-terminal residues level, we solved the structure of the 25 kDa IFNAR2-EC 203–212. b Each hydrogen bond was defined by a pair of distance restraints, by multidimensional NMR. This first solution structure ˚ ˚ dNH-O Ͻ 2.05 A and dN-O Ͻ 2.9 A. of a hematopoietic receptor establishes an atypical per- c Distance root-mean-square deviations (rmsd) are from the upper pendicular orientation and a well-structured hinge re- limit of the restraint and are calculated relative to the mean structure. gion between the two fibronectin domains of unbound d Idealized geometries are based upon CNS parameters (parallhdg. IFNAR2. The IFNAR2-EC/IFN␣2 complex was then mod- pro). eled with the structures of IFNAR2-EC solved in the present study and IFN␣2 solved previously by NMR (Klaus et al., 1997). The docking of IFN into the IFNAR2- fibronectin domains are characterized by seven ␤ EC binding site is based on four intermolecular con- strands arranged in a ␤ sandwich, made of sheets ABE straints derived from double mutant cycle (DMC) mea- and CCЈFG. Two disulfide bonds found in IFNAR2-EC, surements (Roisman et al., 2001). Our findings reveal RC58-RC66 and RC180-RC200, are conserved in both that the receptor binding site is formed by an extensive IFN␥R and TF. A third disulfide bond, RC12-RC95, con- and predominantly aliphatic hydrophobic patch, which necting the two ␤ sheets of the N domain, is unique to interacts with a matching hydrophobic surface of IFN␣2. IFNAR2-EC. Both domains are stabilized at either end by An adjacent pattern of charged residues of alternating a hydrophobic core, usually formed around a conserved sign guides the ligand into its binding site. These results aromatic residue. IFNAR2-EC exhibits such hydropho- account for the crossreactivity of IFNAR2-EC, as well bic cores around residues RW27, RY79, RF127, and RY179. as the lack of involvement of its C domain in the binding Sequence homology between IFNAR2-EC and other site, which distinguishes IFNAR2-EC from other HCRs. HCRs is limited. Despite several differences in sequence and structure, IFNAR2-EC most resembles the receptor for IFN␥. A superposition of the fibronectin N domains Results of the two receptors reveals 65 equivalent C␣ atoms (out of a possible 82) with an rmsd of 1.8 A˚ , while their Overall Structure of IFNAR2-EC C domains (66 equivalent C␣ atoms out of a possible The structure of IFNAR2-EC was determined at a root- 85) superimpose at 1.9 A˚ rmsd. Notable differences be- mean-square deviation (rmsd) of 0.67 A˚ for backbone tween the two receptors are the additional ␤ bulge on atoms (Table 1). Over 95% of residues 11–203 appear the GN ␤ strand, longer BCC loop, and additional C181- in favorable regions of a Ramachandran plot. Terminal C186 disulfide bond present in IFN␥R. (␤ strand lettering residues 1–11 and 206–212 are unstructured, as indi- follows immunoglobulin nomenclature, domains are de- cated by the lack of long-range NOE contacts for this noted by a subscript N or C, and loops are designated by region. Residues 13–99 of the N-domain and residues their flanking ␤ strands.) Although class II HCRs typically 111–203 of the C domain form two fibronectin type III lack the “WSXWS” box appearing in class I receptors modules, connected by a linker segment (residues 100– (Bazan, 1990), residues 90–94 of IFNAR2-EC (TTLFS)

110), which includes a typical ␣ helix (Figure 1). Both preceding the GN strand adopt its characteristic polypro- Molecular Basis of IFN Receptor Ligand Binding 793

Figure 1. Solution Structure of IFNAR2-EC ␤ strands and ␣ helices are shown in blue and red, respectively. N- and C-fibronectin domains referred to in the text are denoted. Residues 1–9 of the N terminus and 207–212 of the C terminus are omitted for clarity. (A) Stereo view of an ensemble of 22 superim- posed lowest-energy NMR-derived back- bone traces of IFNAR2-EC generated with the program MOLMOL (Koradi et al., 1996). (B) Ribbon representation of the average min- imized structure of IFNAR2-EC generated by InsightII (Accelrys).

line type II backbone conformation. This ␤ bulge corre- region. In light of the perpendicular domain orientation sponds to homologous segments in IFN␥R (residues that we observe for IFNAR2-EC, classification of HCRs 93–97, ESAYA) and IL-10R1 (residues 87–91, HSQWT). according to interdomain angle must be reassessed.

However, the highly basic motif of the FN ␤ strand in- This perpendicular orientation is dictated by interac- volved in the characteristic cation-␲ stacking interaction tions of hinge region residues RS106–RE108 with both is conspicuously absent in IFNAR2-EC. Overall, while domains (Figure 2). These residues form a ␤ strand paral- sharing the common double fibronectin domain fold, lel to the AN ␤ strand, a motif conserved in both IFN␥R IFNAR2-EC exhibits significant structural differences (Walter et al., 1995) and IL-10R1 (Josephson et al., 2001). when compared to the family of HCRs. The contact between the hinge region and the C domain is stabilized by hydrophobic interactions of RF107 with R R Interdomain Orientation of IFNAR2-EC residue L139 from the BCC loop and side chain interac- R R R R A striking feature of IFNAR2-EC is the mutually perpen- tions of S106 and E108 with residues I194 and S196 R dicular orientation of its two domains. Class II HCRs from the FGC loop. The adjacent Pro-Pro motif (residues were previously believed to exhibit an interdomain angle 109–110) further rigidifies the interdomain hinge region 120Њ, on the basis of the X-ray structures of IFN␥R and joins the two parallel ␤ strands in stabilizing theف of (Walter et al., 1995) and the related TF (Harlos et al., perpendicular interdomain orientation. Furthermore, ␤ R R 1994). However, in IL-10R1, a recently reported X-ray side chains of the ABN turn (residues N20– F21) pro- structure of this family, the domains are perpendicularly trude from the N-domain and interact with the turn oriented (Josephson et al., 2001). Our structure indicates formed by residues RP128–RI130 of the C domain. 25Њ The perpendicular orientation of IFNAR2-EC domainsف that the C domain of IFN␥R must be rotated by about an axis passing through the hinge region to super- affects the receptor surface presented for ligand bind- R R impose upon the analogous domain in IFNAR2-EC. Sig- ing. The BCC (residues 131–139) and FGC (residues 1 nificantly, measurement of DNH residual dipolar cou- 187–194) loops are retracted from the vicinity of the plings (RDCs) for each of its domains (data not shown) ligand binding site, while these loops are active partici- independently validates the interdomain orientation ob- pants in the binding site of other HCRs (Clackson et al., tained from the NOE data. Moreover, the observation of 1998; Josephson et al., 2001; Walter et al., 1995). This significant RDCs testifies to the rigidity of the hinge is in keeping with the lack of involvement of the IFNAR2 Structure 794

The large hydrophobic surface is surrounded by a ring of polar and charged residues, including RE84, RS96, RH97, RN98, RD104, RD138, and RD189, which delimit the hydrophobic patch. The right side of this ring exhibits a distinct pattern of charged residues of alternating sign, comprising residues RE50, RK48, and RE77 and polar residue RH76. A hydrophobic core formed on the oppo- site face of the ␤ sheet around aromatic residues RY43, RW72, and RY79 interacts with residues RI45, RL68, RF99, and RL101, thus stabilizing the entire binding site structure.

Modeling of the IFNAR2-EC/IFN␣2 Complex with Double Mutant Cycle Constraints It is generally accepted that a small number of experi- mentally obtained intermolecular constraints is suffi- cient to predict, with reasonable accuracy, the complex between two proteins, provided no major structural changes accompany its formation (Clore, 2000). This condition is satisfied for IFNAR2-EC, since changes in chemical shifts induced by IFN␣2 binding were limited to the combining site region (Chill et al., 2002). Double mutant cycle (DMC) measurements have been used to correlate the contributions of specific residues to the binding affinity in protein-protein interactions (Schreiber and Fersht, 1995). These correlations, when interpreted as interresidue spatial proximities, have been used to predict the structure of several protein-protein com- Figure 2. Detailed View of the Interdomain Interface of IFNAR2-EC plexes (Ackermann et al., 1998; Zvi et al., 2000). We A network of interactions between residues of two fibronectin do- therefore used the four most significant previously iden- 90Њ interdomain angle upon tified DMC constraints (Roisman et al., 2001; shownف mains and the hinge region confers an IFNAR2-EC. Backbone segments are depicted as blue ribbons and in Table 2) to drive a simulated-annealing rigid-body labeled as in the text. Side chains contributing to these interactions docking algorithm, thereby modeling the IFNAR2-EC/ are shown as sticks. The backbone of the BCC loop (containing residue L139) is omitted for clarity. IFN␣2 complex. A fifth constraint, incompatible with the structure of IFNAR2-EC, was omitted from the calcula- tions. This docking algorithm yields a complex with an C domain in ligand binding. In addition, the exposure rmsd of 1.0 A˚ for the superposition of all backbone atoms of the interdomain hinge ␣ helix is increased, accounting in structured segments of the N domain of IFNAR2-EC for the importance of residues RW100, RI103, and RD104 and IFN␣2 (Figure 4A). for ligand affinity. Analysis of the obtained complex shows that over 800 A˚ 2 of IFNAR2-EC are buried upon ligand binding, Structure of the IFNAR2-EC Binding Site 75% of which are contributed by residues RM46, RS47, Our previous analysis of the changes in IFNAR2-EC RP49, RH76, RE77, RW100, and RI103. All three segments chemical shifts induced upon IFN␣2 binding mapped of the binding site closely contact the IFN␣2 ligand (Fig- R the ligand binding site to a contiguous surface upon the ure 4B). The CCЈN loop protrudes into a groove formed N domain of the receptor (Chill et al., 2002). Juxtaposing by the ␣AB loop and ␣E helix, contributing over 50% R these changes upon the structure of IFNAR2-EC locates of the decrease in surface accessibility. The EFN loop the receptor binding site to the anterior face of the contributes approximately 25% of the surface buried R Ј ␤ ␣ CC FGN sheet (Figure 3A). Residues contributing to upon binding, interacting with the E helix. The receptor R ␣ the binding site are located on the CCЈN (residues 44–52) interdomain helix interacts with the A helix of the ligand R and EFN (residues 74–82) loops and the ␣ helix in the and forms 20% of the buried surface. In the binding interdomain hinge (residues 100–104). The most striking surface on IFN␣2, the ␣A helix, ␣AB loop, and ␣E helix feature of the IFNAR2-EC binding surface, which mea- contribute 20%, 30%, and 45%, respectively, of the sures approximately 23 ϫ 20 A˚ 2, is a motif of parallel 650 A˚ 2 buried upon binding. The change in surface ac- hydrophobic and hydrophilic striations. Residues RM46, cessibility in IFN␣2 is evenly distributed throughout the RP49, RL52, RV80, RV82, RW100, and RI103 form an elon- binding site, with only one residue (␣R149) contributing gated hydrophobic patch that spans the binding site. over 10% of the buried surface, whereas IFNAR2-EC Of interest is the relative absence of aromatic residues has five such “hotspot” residues. in the IFNAR2-EC binding site, with W100 as the sole exception. In contrast, this large hydrophobic patch is The IFNAR2-EC/IFN␣2 Complex Reveals a Highly absent, and aromatic, rather than aliphatic, side chains Compatible Binding Interface make major contributions to ligand binding in the closely A detailed analysis of the binding interface is facilitated related IFN␥R, IL-10R1, and GHR (Figures 3B–3D). by an open-book representation of the IFNAR2-EC/ Molecular Basis of IFN Receptor Ligand Binding 795

Figure 3. The IFNAR2-EC Binding Site and Comparison to Homologous Helical Cytokine Receptors Binding site residues are labeled and colored as follows: green, aliphatic; dark green, aromatic, red and blue, negatively and positively charged residues, respectively; light blue, histidine; cyan, asparagine and glutamine; orange, serine and threonine. (A) The IFNAR2-EC binding site and surroundings, demonstrating the hydrophobic and hydrophilic striations and the pattern of charge distribution. The binding site (as defined by ligand-induced chemical shift changes) is delimited in black. (B–D) Binding sites of three helical cytokine receptors, (B) IFN␥R, (C) IL-10R1, and (D) GHR. Binding sites are defined by loss of surface accessibility upon complex formation. Receptors are rotated to facilitate viewing of the binding sites.

IFN␣2 complex (Figure 5A). The striated motif observed sumed to be unprotonated at a pH above 7.5 (where for the IFNAR2-EC binding surface interacts with a binding is strongest) and, as suggested by the pH de- highly complementary array of hydrophobic and hydro- pendence of the dissociation of the complex (Piehler philic patches on the binding surface of IFN␣2. Residues and Schreiber, 1999a), serves as an acceptor for the side RE50, RK48, and RE77 of the hydrophilic strip interact chain hydroxyl proton of ␣S152. The resulting overall with a matching array of alternating charges upon IFN␣2 pattern of four intermolecular electrostatic interactions formed by residues ␣R33, ␣D35, and ␣R149. RH76 is as- of alternating polarity on the surface of both receptor and ligand is particularly striking (Figure 5B). At the heart of the binding interface are the two complementary hy- Table 2. Intermolecular Constraints Used in Modeling the drophobic strips, with receptor residues RM46, RP49, IFNAR2-EC/IFN␣2 Complexa,b RV80, RV82, RW100, and RI103 of the anterior face of the R Ј ␤ ␣ ␣ IFNAR2-EC IFN␣2 CC FGN sheet interacting with IFN 2 residues M16 ␣ ␣ ␣ R c ␣ and A19 of the A helix, L26 and L30 of the AB loop, and M46 R144 ␣ ␣ R ␰d ␣ ␦d A145 and M148 of the E helix. The guanido moieties K48N D35O ␣ ␣ RH76N␦/N⑀d ␣S152O␥ of both arginine residues R22 and R144 are oriented ␣ RE77O⑀d ␣R149N␩d inward, interacting in the IFN␣2 core with E141. Their polymethylene side chains may therefore further con- a Constraints derived from double mutant cycle analysis (Roisman et al., 2001). All constraints used satisfied the condition |⌬⌬G| Ͼ tribute to the hydrophobic contact area. The role played 3.0 kJ/mol. Of the five constraints originally used, the least signifi- by the hydrophilic ring of residues surrounding the bind- cant one (RY43/␣F27) was incompatible with the obtained structures ing site remains unclear. and was not included in the docking algorithm. b Intermolecular constraints were triply weighted compared with in- Binding Site Electrostatics tramolecular constraints during the docking algorithm. c The hydrophobic interaction between RM46 and ␣R144 was inter- A curious feature of the binding site is a disproportionate preted as an ambiguous distance constraint of 2.9 A˚ between side presence of histidine residues, including RH76, which chain protons. actively participates in binding, and RH97 and RH187, d Electrostatic interactions were interpreted as a distance of 2.9 A˚ which border the binding site. These residues cause a between donor and acceptor atoms. Ambiguous constraints were dramatic change in electrostatic potential of the utilized when appropriate. IFNAR2-EC binding surface as a function of pH, from a Structure 796

between samples at pH 8 (at which all NMR measure- ments were conducted) and pH 7.2 and visible precipita- tion at pH 6.7. While surprising for a protein with 14 excess negative charges, this behavior could be ex- plained by the pH-induced increase in hydrophobic character of the binding surface on IFNAR2-EC. Thus, the overall electrostatic character of the binding surface of IFNAR2-EC may be responsible for the tight complex it forms with IFN ligands. Both the IFN␥ (Fischer et al., 1988) and the type I IFN signaling complexes (Zoon et al., 1983) have been shown to undergo ligand-dependent endocytosis and then complex dissociation and cellular degradation of the ligand. The acidic endosomal pH has been impli- cated in the dissociation of many such signaling com- plexes (Mukherjee et al., 1997), in some cases by trig- gering a conformational change in the complex (DiPaola and Maxfield, 1984). The histidine cluster in the IFNAR2- EC binding site may provide the molecular mechanism for dissociation of the internalized IFNAR2/IFN␣2 com- plex. Protonation of the histidine residues would likely destabilize the complex by repelling the IFN␣2 binding surface. Further experiments are necessary to explore the role of electrostatics in the cellular recycling of the type I IFN receptor.

Discussion

Structural Comparison with Other Helical Cytokine Receptors The crystal structures of TF (Harlos et al., 1994) and two class II HCR ectodomains, IFN␥R (Walter et al., 1995) and IL-10R1 (Josephson et al., 2001), have been pre- viously determined. While sharing the same double fi- bronectin domain fold, these four proteins exhibit low levels of sequence homology. A comparison of IFNAR2- EC to the other proteins reveals 7–12 insertions or dele- tions, involving over 15% of their sequences. The vast majority of these occur in loop segments, including dele- tions in the loops that form the binding site. These fac- tors have severely undermined previous attempts to Figure 4. The IFNAR2-EC/IFN␣2 Complex model the structure of IFNAR2-EC. Two previously pro- (A) Average ribbon structure of the complex between IFNAR2-EC posed models for IFNAR2-EC, one based on sequence (blue) and IFN␣2 (red). ␤ strands of IFNAR2-EC and helices of IFN␣2 homology (Piehler and Schreiber, 1999b) and the other are labeled. Regions of IFN␣2 involved in binding to the receptor improved with NMR-derived secondary structure data are highlighted in orange. (Chill et al., 2002), demonstrate this clearly. Sequence ␣ (B) Detailed representation of the IFNAR2-EC/IFN 2 interface. The homology is particularly low in the C domain, and model- N domain CCЈFG ␤ sheet is shown as blue ribbons, and the receptor side chains interacting with IFN␣2 are depicted as sticks. The sur- ing fares poorly, as the models superimpose upon the face of IFN␣2 domains interacting with the receptor and the underly- NMR structure at 6.6 and 3.7 A˚ , respectively. Further- ing backbone trace are shown in orange. Receptor ␤ strands, IFN␣2 more, a superposition of the 23 residues of the NMR- helices, and highlighted residues are labeled. determined binding site yields rmsd values of 3.3 and 2.6 A˚ between the models and the solution structure. The main source of structural variance in the binding R highly negatively charged surface at pH 8 (unprotonated site, the CCЈN loop, is highly detrimental to models of histidines) to nearly uncharged at pH 5 (protonated histi- the IFNAR2-EC/IFN␣2 complex, since this loop is also dines) (Figure 6). This change is unparalleled in other the major contributor to the binding interface. The diffi- regions of the receptor. The affinity of IFNAR2-EC to culties in modeling IFNAR2-EC underline the impact of IFN␣2 has been shown to be maximal at a pH above the structure presented in this study upon our under- 7.5, and its significant decrease at slightly acidic pH standing of IFNAR2-EC and its binding to IFNs. values has been attributed to a histidine residue with an elevated pKa of 6.7 (Piehler and Schreiber, 1999b). The Structure of the Binding Interface Accounts We have observed a strong pH dependence of the solu- for Mutagenesis Results bility of free IFNAR2-EC, manifested as a 25%–35% Mutations of residues involved in the IFN␣2 binding site decrease in amide proton transverse relaxation times would be expected to compromise its structure, thus Molecular Basis of IFN Receptor Ligand Binding 797

Figure 5. Open Book Representation of the IFNAR2-EC/IFN␣2 Binding Site (A) Binding surfaces of IFNAR2-EC (left) and IFN␣2 (right) are portrayed in spacefill mode. Highlighted IFNAR2-EC residues exhibit IFN␣2- induced changes in chemical shift. Highlighted IFN␣2 residues lose over 10% of accessible surface upon binding. Residues are colored as in Figure 3. Thin lines connect residues with double mutant cycle energies above 3 kJ (Roisman et al., 2001) and involved in constraints used in the docking algorithm. (B) Schematic representation of the binding sites demonstrating their compatibility. hindering the interaction between IFNAR2-EC and of the binding site, thereby decreasing IFN␣2 affinity IFN␣2. Analysis of the binding site structure presented (Chuntharapai et al., 1999; Lewerenz et al., 1998; Piehler herein reveals a good correlation with experimental mu- and Schreiber, 1999b) (Figure 7A). Mutations of polar tagenesis data. The hydrophobic strip of the binding residues, including RK48A, RE50A, RH76A, and RE77A, surface coincides with the largest hydrophobic patch in adversely affect IFN␣2 binding and reduce the comple- IFNAR2-EC (Piehler et al., 2000). The fact that the RM46A mentarity between ligand and receptor. mutation is the single mutation most detrimental to The model of the IFNAR2-EC/IFN␣2 complex identi- IFN␣2 binding is consistent with its central position in fies a second category of mutations that reduce ligand this binding surface. Similarly, the RV80A, RW100A, and affinity by affecting residues lacking direct contact with RI103A mutations significantly lower the hydrophobicity the ligand, but playing an active role in stabilizing the binding site conformation. The IFN␣2 affinity of the RT44A mutant drops significantly because the absence of the ␥-methyl group destabilizes the conformation of R the CCЈN loop fold (Figure 7A). Four mutations of resi- R dues from the opposite face of the CCЈFGN ␤ sheet, RY43A, RI45A, RW72A, and RY79A, do not directly contact IFN␣2 yet exert moderate to strong effects upon IFN␣2 affinity (Lewerenz et al., 1998; Piehler and Schreiber, 1999b). Residues RY43 and RY79 pack tightly against the indole and benzene ring of RW72, respectively, stabiliz- R R ing the interface between the CCЈN and EFN binding loops with this unique aromatic stacking pattern. RI45 further contributes to this hydrophobic core by inter- acting with the opposite face of the RY79 aromatic ring (Figure 7B). This structural effect is clearly demonstrated by the RW72A and RY79A mutations, which have been shown to produce poorly folded and misfolded proteins, respectively (Piehler and Schreiber, 1999b). Figure 6. Electrostatics of the IFNAR2-EC Binding Site A similar agreement is observed between mutagene- The electrostatic potential for the surface of the IFNAR2-EC binding sis data for the binding surface on IFN␣2 and our model site was calculated with the Delphi module of InsightII (Accelrys). ␣ The surface is shown at pH 5 (left) and pH 8 (right). A pKa value of of the binding interface. Of the 12 IFN 2 residues shown 6.0 was assumed for histidine residues. Blue, gray, and red shades to significantly (more than 4-fold) decrease IFNAR2-EC ␣ represent positive, neutral, and negative potentials, respectively. binding (Piehler et al., 2000), 11 (all but L15) are located Structure 798

Figure 7. Detailed View of the Binding Site of IFNAR2-EC (A) Frontal and (B) reverse views of the N domain CCЈFG ␤ sheet. Both views highlight residues contributing to the hydrophobic strip of the binding surface and residues forming its stabilizing hydrophobic scaffold. The four ␤ strands are depicted as blue ribbons and labeled appropriately.

in our NMR-determined binding site, including the six to the formation of their complexes with IFN␥ and IL- “hotspot” residues ␣L30, ␣R33, ␣R144, ␣A145, ␣M148, 10, respectively. and ␣R149. These residues are also conserved through- In binding all IFN␣ isotypes as well as IFN␤, IFNAR2- out all IFN␣ subtypes. ␣L15 is mostly buried in the struc- EC exhibits a crossreactivity that distinguishes it from ture of IFN␣2, and its adverse effect upon complex for- other HCRs. This may be explained by the unique char- mation may reflect a structural role in stabilizing the A acter of its binding site. While aromatic residues typically helix. The hydrophobic strip in our IFN␣2 binding surface form specific interactions with stringent spatial require- extends down the A helix to include residues ␣M16 and ments, the aliphatic side chains of the IFNAR2-EC bind- ␣A19. IFN␣ isotypes exhibit only conservative substitu- ing site may present a more malleable hydrophobic sur- tions for these solvent-exposed residues, supporting face, capable of accommodating the corresponding their involvement in the binding site. The methylene side hydrophobic patch of different IFN ligands. In the ab- chain of ␣R22 may further contribute to this hydrophobic sence of aromatic side chains, correct spatial position- surface, although its low level of conservation indicates ing of the ligand may be maintained by the pattern of that it does not play a critical role in binding to the charge distribution surrounding the hydrophobic patch. receptor. The molecular view of the IFN binding mode Such electrostatic guiding is in keeping with the pre- presented here provides, therefore, a comprehensive viously described paradigm for protein-protein interac- structural context for the interpretation of in vitro and tions (Schreiber and Fersht, 1996; Selzer and Schreiber, cellular binding assays. 2001). Interestingly, the single aromatic side chain in the binding site (RW100) may be responsible for specific Implications of Structure on Receptor recognition of IFN␤. IFN␤ activity is abolished in the Crossreactivity and Ligand Selectivity RW100A mutant, while IFN␣ activity decreases only A comparison between the binding sites of IFNAR2- 4-fold. The replacement of ␣A19 of the A helix by ␤W22 EC and other HCRs reveals two important differences in IFN␤ may allow a favorable interaction between the (Figure 3). The large hydrophobic patch spanning the aromatic rings of these two tryptophan residues (Piehler binding site in IFNAR2-EC is not observed in IFN␥R, IL- et al., 2000). Furthermore, a second nonconservative 10R1, or GHR, whose binding site is typically a mosaic substitution observed in the A helix of IFN␤, ␤K19 replac- of hydrophobic and polar residues involved in intermo- ing ␣M16, may join important sequence variations on lecular contacts. Furthermore, these three counterparts the ␣C helix in contributing to specific ligand recognition of IFNAR2-EC are characterized by major contributions by IFNAR2-EC. of aromatic residues to the binding energy, while the A comparison of D2O exchange rates between free hydrophobic patch of IFNAR2-EC is predominantly ali- and complexed IFNAR2-EC demonstrated that ligand phatic. In the GHR binding site, residues W104 and W169 binding is accompanied by an overall tightening of the of lost surface accessibility N domain, but not the C domain. Notably, this effect %70ف alone encompass upon ligand binding (Clackson et al., 1998; de Vos et was not confined to the IFN␣2 binding site (Chill et al., al., 1992). Similarly, residues Y49, W56, W82, and W207 2002). This conformational flexibility of the N domain of IFN␥R and Y43, W48, and Y75 of IL-10R1 are essential may enhance the capability of the receptor to accommo- Molecular Basis of IFN Receptor Ligand Binding 799

date different ligands in the receptor binding site. The experimental data on the precise role of each binding ligand-induced tightening of the N domain may then be domain in the signaling complex, it lays the foundations involved in subsequent events of the signaling cascade. for further structural studies. In summary, the unique structure and flexible character of the IFNAR2-EC binding site account for its versatility Biological Implications in binding multiple ligands. Two receptor subunits, IFNAR1 and IFNAR2, regulate the activity of type I interferons (IFNs) in humans, thereby mediating a host of cellular processes, including apo- Implications of Interdomain Orientation ptosis and inhibition of viral replication and cell growth upon IFN Signaling (Stark et al., 1998). IFNs have been the focus of extensive Together with the recently determined X-ray structure biomedical research and are widely used to treat various of IL-10R1, the perpendicular interdomain orientation of types of cancer and viral diseases (Deonarin et al., 2002). IFNAR2-EC represents 50% of the known structures of A structural understanding of ligand binding and speci- class II HCRs. This disproves the common belief that the ficity could greatly enhance the efficacy of such 120Њ. The orientationف interdomain angle of this class is IFN-based therapies. Despite extensive efforts, difficul- established for IFNAR2-EC obtained in solution is partic- ties in crystallizing the membranal receptor chains and ularly significant, since, unlike X-ray structures, it is un- low homology within the helical cytokine receptor family affected by crystal packing forces. NMR is well suited have frustrated attempts to obtain information at the for structure determination of multidomain macromole- molecular level for the type I IFN receptor. cules (Barbato et al., 1992), and this capability is The structure of the IFNAR2 ectodomain (IFNAR2- strengthened by the advent of residual dipolar-coupling EC) reveals this molecular basis for IFN binding and measurements (Bax et al., 2001). signaling. The two fibronectin domains of IFNAR2-EC The perpendicular interdomain orientation of IFNAR2- are perpendicularly oriented, explaining the lack of EC increases the exposure of the hinge region for inter- involvement of the C domain in the binding site. This action with IFN and at the same time retracts the C result obtained for the free receptor in solution is particu- domain from the vicinity of the binding site. This explains larly important, since variations in interdomain angle the lack of involvement of the C domain in ligand binding, and domain rigidity have been studied in the context of which is unique to IFNAR2-EC. Neither mutagenesis signaling (Muller et al., 1998). The structure also repre- (Lewerenz et al., 1998; Piehler and Schreiber, 1999b) sents an important step toward determination of the nor NMR (Chill et al., 2002) studies could identify a contri- structures of type I IFN signaling complexes. We have bution of the C domain to IFN binding, and the IFN␣2 demonstrated this by utilizing double mutant cycle- affinities of IFNAR2-EC and a truncated polypeptide in- derived constraints to dock the IFN␣2 ligand into its cluding the N domain and hinge are practically indistin- binding site upon IFNAR2-EC. The predominantly ali- guishable (Roisman et al., 2001). The interdomain orien- phatic hydrophobic binding patch of IFNAR2-EC is ca- tation also resolves difficulties raised by a previous model, pable of binding multiple ligands, guided electrostati- in which the assumed 120Њ interdomain angle created a cally by an adjacent motif of alternating charges. A significant contact area between IFN␣2 and the C domain similar array of hydrophobic and charged strips on (Lewerenz et al., 1998; Piehler and Schreiber, 1999b). IFN␣2 presents a highly complementary binding surface. IFN signaling is contingent upon IFN-induced associa- The structure also identifies receptor residues that as- tion of IFNAR2 and IFNAR1. The extracellular domain sume key roles in stabilizing the binding interface. The of IFNAR1 comprises two linked IFNAR2-EC-like poly- electrostatics of the binding site account for the ob- peptides, D1 and D2, or four fibronectin domains. Al- served pH dependence of binding affinity and may play though the structural factors governing the formation of a role in recycling of the IFN signaling complex. To- this ternary signaling complex have yet to be deter- gether, the structures of IFNAR2-EC and its complex mined, a plausible mode of assembly may be suggested with the IFN␣2 ligand provide a molecular understanding from the results of this study and mutagenesis data of IFN biochemistry, suggest a plausible assembly for on the IFNAR1/IFN␣2 interface (Figure 8). The binding its signaling complex, and lay the groundwork for further surface of IFN␣2, comprising the ␣A and ␣E helices and biological experimentation. In light of the increasing the ␣AB loop, binds to the N domain of IFNAR2, forming therapeutical applicability of IFNs, these findings prom- the binding interface described in this study. The oppo- ise to be particularly valuable for future progress in this site face of IFN␣2, comprising helices ␣B and ␣C, is important field. assumed to interact with IFNAR1. Sequence variations in the ␣C helix between IFN␣ and IFN␤ have been impli- cated in mediating ligand-specific signaling by inter- Experimental Procedures acting differently with the IFNAR1 binding site (Korn et NMR Sample Preparation al., 1994). This binding surface has been located by IFNAR2-EC was expressed in appropriately labeled Celtone medium mutagenesis at the cleft formed by the central two of the (Spectragases) and purified as previously described (Chill et al., four IFNAR1 fibronectin domains (Cutrone and Langer, 2002). All NMR experiments were conducted at concentrations of 2001). The orientation of D1 and D2, each modeled here 0.25–0.5 mM of IFNAR2-EC in 20 mM Tris buffer at pH 8 and 0.02% NaN3. Samples prepared in this manner included unlabeled IFNAR2- by molecular replacement with IFNAR2-EC as a tem- 15 13 15 ␣ ␣ EC, N-IFNAR2-EC in H2O, and C, N-IFNAR2-EC samples in both plate, enables them to interact with the B and C helices 15 H2O and D2O. Weakly aligned samples of N-IFNAR2-EC in H2O of IFN␣2, thereby forming the signaling complex (Figure were prepared in polyacrylamide gel (monomer/crosslinker ratio of 8). While verification of this hypothesis awaits further 37.5:1) swelled with a solution of the protein as previously described Structure 800

Figure 8. Postulated Assembly of the Type I IFN Signaling Complex Stereo view of a suggested organization of the ternary type I IFN signaling complex formed by IFN␣2 and both receptor subunits. The modeled IFNAR2-EC/IFN␣2 complex appears as blue (IFNAR2-EC) and red (IFN␣2) ribbons, with their binding interface highlighted in orange. Each of the two IFNAR1 200-residue domains (D1 and D2) is modeled on the basis of the structure of IFNAR2-EC and shown in surface representation. Regions upon IFNAR1 and IFN␣2 postulated to interact in the signaling complex (Cutrone and Langer, 2001; Deonarin et al., 2002; Runkel et al., 2000) are highlighted in purple. The association of the two receptor subunits triggers the IFN signaling cascade.

(Sass et al., 2000; Tycko et al., 2000). Longitudinal pressure was ment. NOE-derived constraints from each spectrum were calibrated applied on the swelled gel in a Shigemi NMR tube with the insert. with the appropriate proton-proton distances with fixed geometry or of well-defined secondary structure. φ angle constraints were NMR Spectroscopy derived from measured 3J couplings. The φ angle was constrained NMR spectra were acquired at 32–35ЊC on Bruker 500 or 800 MHz to Ϫ65ЊϮ25Њ for 3J values under 6 Hz and to Ϫ120ЊϮ30Њ for 3J NMR spectrometers. Backbone assignment and measurement of values over 8 Hz. Additional angular constraints were obtained from 3 13 JHNH␣ couplings have been previously described (Chill et al., 2002). an analysis of backbone and C␤ chemical shifts with the TALOS 1H and 13C resonances of side chains were assigned by 3D HCCH- program (Cornilescu et al., 1999). Hydrogen bonds were interpreted

COSY (Bax et al., 1990a) and 3D HCCH-total correlation spectros- as a pair of distance constraints, dNH-O Ͻ 2.05 A˚ and dN-O Ͻ 2.9 A˚ . copy (TOCSY) (Bax et al., 1990b), HBHA(CBCACO)NH (Grzesiek and All structure calculations employed version 1.1 of the CNS program Bax, 1993), and (H)CC-TOCSY-(CO)NH (Grzesiek et al., 1993) experi- (Bru¨ nger et al., 1998). The structure of each domain of IFNAR2- ments acquired for a uniformly 15N/13C-labeled sample. Assignment EC (including the hinge segment) was calculated separately with of protons of aromatic side chains was accomplished by 2D distance geometry and simulated annealing. Constraints were intro- CB(CGCD)HD and 2D CB(CGCDCE)HE experiments (Yamazaki et duced into the calculation in an iterative manner, with a previously 15 13 al., 1993) acquired for a uniformly N/ C-labeled sample in D2O and calculated model (Chill et al., 2002) and in-house scripts assisting in assisted by 2D homonuclear TOCSY (31 ms mixing time) and NOESY assignment of ambiguous NOEs. Superposition of the hinge regions (75 ms mixing time) acquired for a 0.7 mM unlabeled sample. Experi- yielded initial structures for the final refinement stage. The final ments used to derive NOE constraints included the aforementioned refinement included several simulated annealing steps and was homonuclear, 3D 15N-separated (Marion et al., 1989), and 3D 13C- based upon 1876 interresidual interproton distance, 134 hydrogen separated (Ikura et al., 1990) NOESY experiments, with mixing times bond, 88 ␸ angle, 92 TALOS-derived backbone dihedral, and 109 of 70–80 ms. [1H,15N]-RDCs were measured by comparing the DSSE- N-H RDC restraints. Structures were displayed with the InsightII [1H,15N]IPAP-HSQC spectrum (Cordier et al., 1999) acquired for iso- software package (Accelrys). tropic and aligned 15N-labeled samples. Data were processed and analyzed with XWINNMR (Bruker) and the NMRPipe/NMRDraw (De- Molecular Docking and Modeling laglio et al., 1995) software packages. IFN␣2 was docked into its binding site upon IFNAR2-EC using four intermolecular distance constraints derived from analysis of double Structure Determination mutant cycle experiments. All IFNAR2-EC backbone and side chain Distance constraints were obtained from 3D 15N- or 13C-separated NOE constraints were upheld in the docking procedure. The back- NOESY experiments as well as the 2D NOESY homonuclear experi- bone trace of IFN␣2 was fixed with 360 intra- and interhelix C␣-C␣ Molecular Basis of IFN Receptor Ligand Binding 801

distance constraints obtained from its previously published NMR Clore, G.M. (2000). Accurate and rapid docking of protein-protein structure ( accession code 1ITF) (Klaus et al., complexes on the basis of intermolecular nuclear Overhauser en- 1997). Intermolecular constraints detailed in Table 2 were triply hancement data and dipolar couplings by rigid body minimization. weighted. In the initial structures the ligand was randomly rotated Proc. Natl. Acad. Sci. USA 97, 9021–9025. ˚ ف and positioned 20 A away from the binding site. A simulated an- Cohen, B., Novick, D., Barak, S., and Rubinstein, M. (1995). Ligand- nealing protocol was then used to model the structure of the induced association of the type I interferon receptor components. ␣ IFNAR2-EC/IFN 2 complex. The D1 and D2 domains of IFNAR1 Mol. Cell. Biol. 15, 4208–4214. were modeled by aligning their sequences with that of IFNAR2-EC Cordier, F., Dingley, A.J., and Grzesiek, S. (1999). A doublet-sepa- and then by backbone coordinate replacement. This process, as rated sensitivity-enhanced HSQC for the determination of scalar well as calculation of surface accessibilities, was performed with and dipolar one-bond J-couplings. J. Biomol. NMR 13, 175–180. the program InsightII (Accelrys). Cornilescu, G., Delaglio, F., and Bax, A. (1999). Protein backbone Acknowledgments angle restraints from searching a database for chemical shift and sequence homology. J. Biomol. NMR 13, 289–302. We thank Dr. Vitali Tugarinov for assistance with NMR experiments Cutrone, E.C., and Langer, J.A. (2001). Identification of critical resi- and Dr. Tali Scherf for spectrometer assistance. We are also in- dues in bovine IFNAR-1 responsible for interferon binding. J. Biol. debted to Dr. Miri Eisenstein for several helpful discussions on mod- Chem. 276, 17140–17148. eling aspects of this work and Prof. Fred Naider for critical reading de Vos, A.M., Ultsch, M., and Kossiakoff, A.A. (1992). Human growth of this manuscript. This work was supported by the National Insti- hormone and extracellular domain of its receptor: crystal structure tutes of Health and the Israel Academy of Sciences (to J.A.). J.A. is of the complex. Science 255, 306–312. the Dr. Joseph and Ruth Owades Professor of Chemistry. Delaglio, F., Grzesiek, S., Vuister, G.W., Zhu, G., Pfeifer, J., and Bax, A. (1995). NMRPipe: a multidimensional spectral processing system Received: February 3, 2003 based on UNIX pipes. J. Biomol. NMR 6, 277–293. Accepted: April 1, 2003 Published: July 1, 2003 Deonarin, R.D., Chan, D.C.M., Platanias, L.C., and Fish, E.N. (2002). Interferon-␣/␤ receptor interactions: a complex story unfolding. References Curr. Pharm. Des. 8, 99–110. DiPaola, M., and Maxfield, F.R. (1984). Conformational changes in Ackermann, E.J., Ang, E.T., Kanter, J.R., Tsigelny, I., and Taylor, P. the receptors for epidermal growth factor and asialoglycoproteins (1998). Identification of pairwise interactions in the alpha-neuro- induced by the mildly acidic pH found in endocytic vesicles. J. Biol. toxin-nicotinic acetylcholine receptor complex through double mu- Chem. 259, 9163–9171. tant cycles. J. Biol. Chem. 273, 10958–10964. Einhorn, S., and Grander, D. (1996). Why do so many cancer patients Barbato, G., Ikura, M., Kay, L.E., Pastor, R.W., and Bax, A. (1992). fail to respond to interferon therapy? J. Interferon Cytokine Res. 16, Backbone dynamics of calmodulin studied by 15N relaxation using 275–281. inverse detected two-dimensional NMR spectroscopy: the central Fischer, D., Novick, D., Orchansky, P., and Rubinstein, M. (1988). helix is flexible. Biochemistry 31, 5269–5278. Two molecular forms of the human IFN-␥ receptor. J. Biol. Chem. Bax, A., Clore, G.M., Driscoll, P.C., Gronenborn, A.M., Ikura, M., and 263, 2632–2637. Kay, L.E. (1990a). Practical aspects of proton-carbon-carbon-proton Grzesiek, S., and Bax, A. (1993). Amino acid type determination in 13 three-dimensional correlation spectroscopy of C-labeled proteins. the sequential assignment process of uniformly 13C/15N enriched J. Magn. Reson. 87, 620–627. proteins. J. Biomol. NMR 3, 185–204. Bax, A., Clore, G.M., and Gronenborn, A.M. (1990b). H-H correlations Grzesiek, S., Anglister, J., and Bax, A. (1993). Correlation of back- 13 via isotropic mixing of C magnetization, a new three-dimensional bone amide and aliphatic side-chain resonances in 13C/15N-enriched 1 13 13 approach for assigning H and C spectra of C-enriched proteins. proteins by isotropic mixing of 13C magnetization. J. Magn. Reson. J. Magn. Reson. 88, 425–431. B 101, 114–119. Bax, A., Kontaxis, G., and Tjandra, N. (2001). Dipolar couplings in Harlos, K., Martin, D.M., O’Brien, D.P., Jones, E.Y., Stuart, D.I., Poli- macromolecular structure determination. Methods Enzymol. 339, karpov, I., Miller, A., Tuddenham, E.G., and Boys, C.W. (1994). Crys- 127–174. tal structure of the extracellular region of human tissue factor. Nature Bazan, J.F. (1990). Structural design and molecular evolution of 370, 662–666. a cytokine receptor superfamily. Proc. Natl. Acad. Sci. USA 87, Ikura, M., Kay, L.E., Tschudin, R., and Bax, A. (1990). Three-dimen- 6934–6938. sional NOESY-HMQC spectroscopy of a 13C-labeled protein. J. Bru¨ nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Magn. Reson. 86, 204–209. Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., Josephson, K., Logsdon, N.J., and Walter, M.R. (2001). Crystal struc- Pannu, N.S., et al. (1998). Crystallography and NMR system: a new ture of the IL-10/IL-10R1 complex reveals a shared receptor binding software suite for macromolecular structure determination. Acta site. Immunity 15, 35–46. Crystallogr. D Biol. Crystallogr. 54, 905–921. Karpusas, M., Nolte, M., Benton, C., Meier, W., Lipscomb, W.N., and Bukowski, R., Tendler, C., Cutler, D., Rose, E., Laughlin, M., and Goelz, S. (1997). The crystal structure of human interferon-␤ at 2.2-A˚ Statkevich, P. (2002). Treating cancer with PEG intron: pharmaki- resolution. Proc. Natl. Acad. Sci. USA 94, 11813–11818. netic profile and dosing guidelines for an improved interferon-␣2b ␣ formulation. Cancer 95, 389–396. Kirkwood, J. (2002). Cancer immunotherapy: the interferon- experi- ence. Semin. Oncol. 29, 18–26. Chill, J.H., Nivasch, R., Levy, R., Albeck, S., Schreiber, G.E., and Anglister, J. (2002). The human interferon receptor: NMR-based Klaus, W., Gsell, B., Labhardt, A.M., Wipf, B., and Senn, H. (1997). modeling of the IFN-␣2 binding site, and observed ligand-induced The three-dimensional high resolution structure of human interferon- ␣ tightening. Biochemistry 41, 3575–3585. 2a determined by heteronuclear NMR spectroscopy in solution. J. Mol. Biol. 274, 661–675. Chuntharapai, A., Gibbs, V., Lu, J., Ow, A., Marsters, S., Ashkenazi, A., De Vos, A., and Jin Kim, K. (1999). Determination of residues Koradi, R., Billeter, M., and Wuthrich, K. (1996). MOLMOL: a program involved in ligand binding and signal transmission in the human for display and analysis of macromolecular structures. J. Mol. Graph. IFN-␣ receptor 2. J. Immunol. 163, 766–773. 14, 51–55. Clackson, T., Ultsch, M.H., Wells, J.A., and de Vos, A.M. (1998). Korn, A.P., Rose, D.R., and Fish, E.N. (1994). Three-dimensional ␣ Structural and functional analysis of the 1:1 growth hormone:recep- model of a human interferon- consensus sequence. J. Interferon tor complex reveals the molecular basis for receptor affinity. J. Mol. Cytokine Res. 14, 1–9. Biol. 277, 1111–1128. Lewerenz, M., Mogensen, K.E., and Uze, G. (1998). Shared receptor Structure 802

components but distinct complexes for ␣- and ␤-interferons. J. Mol. Uze, G., Lutfalla, G., and Gresser, I. (1990). Genetic transfer of a Biol. 282, 585–599. functional human interferon-␣ receptor into mouse cells: cloning Marion, D., Driscoll, P.C., Kay, L.E., Wingfield, P.T., Bax, A., Gro- and expression of its cDNA. Cell 60, 225–234. nenborn, A.M., and Clore, G.M. (1989). Overcoming the overlap prob- Walter, M.R., Windsor, W.T., Nagabhushan, T.L., Lundell, D.J., Lunn, lem in the assignment of 1H NMR spectra of larger proteins by use C.A., Zauodny, P.J., and Narula, S.K. (1995). Crystal structure of a of three-dimensional heteronuclear 1H-15N Hartmann-Hahn-multiple complex between interferon-␥ and its soluble high-affinity receptor. quantum coherence and nuclear Overhauser-multiple quantum co- Nature 376, 230–235. herence spectroscopy: application to interleukin-1␤. Biochemistry Yamazaki, T., Forman-Kay, J.D., and Kay, L.E. (1993). Two-dimen- 28, 6150–6156. sional NMR experiments for correlating 13C␤ and 1H␦/⑀ chemical shifts Mogensen, K.E., Lewerenz, M., Reboul, J., Lutfalla, G., and Uze, of aromatic residues in 13C-labeled proteins via scalar couplings. J. G. (1999). The type I interferon receptor: structure, function, and Am. Chem. Soc. 115, 11054–11055. evolution of a family business. J. Interferon Cytokine Res. 19, 1069– Zoon, K.C., Arnheiter, H., Zur Nedden, D., Fitzgerald, D.J., and Will- 1098. ingham, M.C. (1983). Human interferon-␣ enters cell by receptor- Mukherjee, S., Ghosh, R.N., and Maxfield, F.R. (1997). Endocytosis. mediated endocytosis. Virology 130, 195–203. Physiol. Rev. 77, 759–803. Zvi, A., Tugarinov, V., Faiman, G.A., Horovitz, A., and Anglister, J. Muller, Y.A., Kelley, R.F., and de Vos, A.M. (1998). Hinge bending (2000). A model of a gp120 V3 peptide in complex with an HIV- within the cytokine receptor superfamily revealed by the 2.4 A˚ crystal neutralizing antibody based on NMR and mutant cycle-derived con- structure of the extracellular domain of rabbit tissue factor. Protein straints. Eur. J. Biochem. 267, 767–779. Sci. 7, 1106–1115. Novick, D., Cohen, B., and Rubinstein, M. (1994). The human inter- Accession Numbers feron-␣/␤ receptor: characterization and molecular cloning. Cell 77, 391–400. The coordinates have been deposited in the Protein Data Bank under accession codes 1N6U (ensemble of low-energy structures) Perry, C., and Jarvis, B. (2001). Peginterferon-␣2a: a review of its and 1N6V (average structure). use in the management of chronic hepatitis C. Drugs 61, 2263–2288. Pestka, S., Langer, J.A., Zoon, K.C., and Samuel, C.E. (1987). Inter- ferons and their action. Annu. Rev. Biochem. 56, 727–777. Piehler, J., Roisman, L.C., and Schreiber, G. (2000). New structural and functional aspects of the type I interferon-receptor interaction revealed by comprehensive mutational analysis of the binding inter- face. J. Biol. Chem. 275, 40425–40433. Piehler, J., and Schreiber, G. (1999a). Biophysical analysis of the interaction of human IFNAR2 expressed in E. coli with IFN␣2. J. Mol. Biol. 289, 57–67. Piehler, J., and Schreiber, G. (1999b). Mutational and structural anal- ysis of the binding interface between type I interferons and their receptor IFNAR2. J. Mol. Biol. 294, 223–237. Radhakrishnan, R., Walter, L.J., Hruza, A., Reichert, P., Trotta, P.P., Nagabhushan, T.L., and Walter, M.R. (1996). Zinc mediated dimer of human interferon-␣2b revealed by X-ray crystallography. Struc- ture 4, 1453–1463. Roisman, L.C., Piehler, J., Trosset, J.-Y., Scheraga, H.A., and Schreiber, G. (2001). Structure of the interferon-receptor complex determined by distance constraints from double mutant cycles and flexible docking. Proc. Natl. Acad. Sci. USA 98, 13231–13236. Runkel, L., deDios, C., Karpusas, M., Betzenhauser, M., Muldowney, C., Zafari, M., Benjamin, C.D., Miller, S., Hochman, P.S., and Whitty, A. (2000). Systematic mutational mapping of sites on human inter- feron-␤-1a that are important for receptor binding and functional activity. Biochemistry 39, 2538–2551. Sass, H.-J., Musco, G., Stahl, S., Wingfield, P.T., and Grzesiek, S. (2000). Solution NMR of proteins within polyacrylamide gels: diffu- sional properties and residual alignment by mechanical stress or embedding of oriented purple membranes. J. Biomol. NMR 18, 303–309. Schreiber, G., and Fersht, A.R. (1995). Energetics of protein-protein interactions: analysis of the barnase-barstar interface by single mu- tations and double mutant cycles. J. Mol. Biol. 248, 478–486. Schreiber, G., and Fersht, A.R. (1996). Rapid, electrostatically as- sisted association of proteins. Nat. Struct. Biol. 3, 427–431. Selzer, T., and Schreiber, G. (2001). New insights into the mechanism of protein-protein association. Proteins 45, 190–198. Stark, G.R., Kerr, I.M., Williams, B.R., Silverman, R.H., and Schreiber, R.D. (1998). How cells respond to interferons. Annu. Rev. Biochem. 67, 227–264. Tycko, R., Blanco, F.J., and Ishii, Y. (2000). Alignment of biopolymers in strained gels: a new way to create detectable dipole-dipole cou- plings in high-resolution biomolecular NMR. J. Am. Chem. Soc. 122, 9340–9341.