Three Loops of the Common γ Chain Ectodomain Required for the Binding of Interleukin-2 and Interleukin-7*

The common γ chain (γc), a subunit of the interleukin (IL)-2, IL-4, IL-7, IL-9, and IL-15 receptors, contributes to both cytokine binding and subsequent signal transduction. Using a model-based site-directed mutagenesis strategy, we have identified residues of the mouse γc extracellular domain that are required for normal γc-dependent enhancement of IL-2 and IL-7 binding. One of these sites, Tyr-103, is homologous to key ligand-interacting residues in the growth hormone and erythropoietin receptors, whereas Cys-161, Cys-210, and Gly-211 may function indirectly by maintaining the functional conformation of γc via formation of an intramolecular disulfide bond. These two cysteines are also required for the integrity of a putative epitope recognized by TUGm2, an antagonistic monoclonal antibody that blocks γc-dependent cytokine binding and bioactivity. These results are consistent with the involvement of three predicted loops in γc that contribute to the binding of both IL-2 and IL-7. Mutations in these loops have also been noted in the γc gene of patients with X-linked severe combined immunodeficiency.

The common ␥ chain (␥c) 1 is a member of the type I cytokine receptor superfamily and functions as a shared subunit of the receptors for IL-2, IL-4, IL-7, IL-9, and IL-15 (reviewed in Refs. 1 and 2). In each of these receptors ␥c regulates signaling via the JAK3 tyrosine kinase associated with its cytoplasmic tail. The biological importance of ␥c is most dramatically illustrated by the fact that mutations of either ␥c or JAK3 are the primary causes of human X-linked severe combined immunodeficiency (X-SCID), characterized by a failure in T and natural killer cell development (3). Although ␥c does not detectably bind cytokines by itself, its recruitment into the IL-2, IL-4, and IL-7 receptor complexes enhances ligand binding. This is most evident in the context of the IL-2 receptor (IL-2R) where ␤␥ heterodimers bind IL-2 with 100-fold higher affinity than IL-2R␤ (4), and the affinity of ␣␤␥ trimers is 10-fold increased over ␣␤ complexes (5). A similar but more modest 3-5-fold increase in affinity has also been observed in the IL-4R (6) and IL-7R (7).
Cytokines binding to type I cytokine receptors are typically folded as bundles of four ␣-helices. The receptor subunits, including ␥c, are characterized by extracellular homology regions composed of two fibronectin type III domains that are each composed of seven ␤-strands. The first domain contains four highly conserved cysteine residues linked by disulfide bonds whereas the second domain contains a highly conserved WSXWS motif (8). The prototypical members of the type I cytokine receptor superfamily are the growth hormone and erythropoietin receptors (EPORs) for which the crystal structure has been determined for the extracellular portion of the receptors together with their ligands (9,10). The structure of the growth hormone receptor (GHR) complex (9) has been used as a template for homology modeling of a number of receptors within the family. Subsequent use of such models in designing structure-function studies has led to the identification of potential receptor-ligand contact sites within the granulocytemacrophage colony stimulating factor receptor (CSFR) (11), granulocyte CSFR (12), and gp130 (13). Several hypothetical models of ␥c-containing complexes have also been reported for the IL-2R, IL-4R, and IL-7R (14 -16). In the present study we used these models as guides for site-directed mutagenesis of putative ligand-interacting regions of the mouse (m) ␥c ectodomain to identify specific amino acids that contribute to ␥c-dependent ligand binding in the context of the heterodimeric IL-2R and IL-7R. Our data suggest that IL-2 and IL-7 binding are dependent on overlapping regions of the m␥c ectodomain that include at least three distinct putative loop segments of the ␥c structure.

EXPERIMENTAL PROCEDURES
Mutagenesis of m␥c-Full-length cDNAs for the m␥c, hIL-2R␤, and mIL-7R␣ were cloned into the pSI mammalian expression vector (Promega, Madison, WI). Site-directed mutagenesis was performed by the Chameleon or Quikchange methods (Stratagene, La Jolla, CA) using 5-10 ng of pSI-m␥c vector as template. The ⌬(4 -34) deletion mutant was prepared by replacing a SacI-NotI fragment of the pSI-m␥c vector (bases 74 -1110 of the m␥c coding sequence) with a shorter piece generated by polymerase chain reaction (bases 168 -1110 with added SacI and NotI linkers). All mutants were verified by DNA sequencing. Mutations were designated by the one-letter code and number of the wild-type (WT) residue in the mature m␥c sequence followed by the code for the substituted amino acid.
Cell Culture and Transient Transfections-COS7 monkey kidney cells were maintained in RPMI 1640 medium containing 5% fetal calf serum, 100 units/ml penicillin, 100 g/ml streptomycin, 30 g/ml Lglutamine, and 50 M 2-mercaptoethanol (complete medium). For transient transfections the cells were harvested by trypsinization, washed, and electroporated in complete medium with 15 mM HEPES at 215 V and 330 microfarads using a Cell-Porator (Life Technologies, Inc.). Mock transfected cells were COS7 electroporated without DNA. In most experiments COS7 cells were co-transfected with two receptor-encoding plasmids (4 g each) such that the final DNA concentration was 20 g/ml. When only one receptor subunit was transfected, or when lower amounts of pSI-m␥c (0.13-4 g) were used to construct standard curves, the total DNA was adjusted to 20 g/ml using empty pSI plasmid. Three days after transfection the cells were harvested with phosphate-buff-ered saline containing 5 mM EDTA, washed with Hanks' balanced salt solution, and used in various assays.
Biochemical Analysis of m␥c-Transfected COS7 cells were extracted in 0.5% Nonidet P-40 in the presence of protease inhibitors as described previously (17). Unfractionated extracts from 10 6 cells/lane were subjected to Western blotting after 10% SDS-polyacrylamide gel electrophoresis under reducing conditions. To detect m␥c, the blots were first incubated with M-20 rabbit antiserum to the cytoplasmic tail of m␥c (1:400, Santa Cruz Biotechnology, Santa Cruz, CA) followed by horseradish peroxidase-conjugated donkey anti-rabbit Ig (1:5000, Amersham Pharmacia Biotech). Bands were visualized by the ECL method (Amersham Pharmacia Biotech). Densitometry was performed using the Scion Image software (Scion Corporation, Frederick, MD).
Flow Cytometry Analysis-Fluorescence-activated cell sorter (FACS) analysis was performed as described previously (18) using a FACScan II flow cytometer and CellQuest analysis software (BD Biosciences, San Jose, CA). Biotinylated anti-m␥c (4G3, 3E12 (19), and TUGm2 (20), BD Biosciences-PharMingen, San Diego CA), biotinylated anti-hIL-2R␤ (Mik-␤3, PharMingen), or anti-mIL-7R␣ (A7R34, kindly provided by S. Nishikawa, Kyoto University, Japan) (21) was used as a primary antibody. Phycoerythrin-streptavidin (PharMingen) or fluorescein isothiocyanate anti-rat Ig (Pierce) was used as a secondary reagent. Data were collected on 5-10 ϫ 10 3 live cells as determined by a combination of forward and side scatter. Because gene expression in COS7 was very heterogeneous, mAb binding (AB) was quantified by considering both the percentage of positively stained cells (%C) and the mean fluorescence intensity (MFI) of the cells. The following formula was used: AB ϭ (%C) ϫ (MFI). Specific mAb binding to COS7-hIL-2R␤/m␥c or COS7-mIL-7R␣/m␥c cells was calculated by subtraction of the AB of similarly stained mock transfected COS7 cells. To determine normalized specific mAb binding (NSAB) for each anti-m␥c mAb, the specific mAb binding was corrected for differences in transfection efficiency (typically Յ15% among cells in the same experiment) based on the expression of the co-transfected hIL-2R␤ or mIL-7R␣. Finally, the relative NSAB to mutant m␥c variants was calculated as 100 ϫ (NSAB ␥cMutant /NSAB ␥cWT ). ␥c surface expression was estimated based on the highest anti-m␥c mAb binding observed for each mutant.
Radioligand Binding Assays-Recombinant human IL-2 and IL-7 (PeproTech, Rocky Hill, NJ) were radiolabeled with 125 I using IODO-GEN tubes (Pierce) to 15-58 Ci/g. Transfected COS7 cells (10 6 /tube) were incubated for 2-4 h at 4°C with 1 nM radioligand in complete medium containing 0.2% NaN 3 and 15 mM HEPES under continuous mixing. The samples were then centrifuged, and cell-associated radioactivity was measured in a ␥-counter. ␥c-specific binding for IL-2 was considered as the counts/min associated with COS7-hIL-2R␤/m␥c transfectants minus the counts/min associated with control cells, i.e. COS7-hIL-2R␤, mock transfected COS7, or cells binding in the presence of 200-fold molar excess of unlabeled ligand, which all gave comparable results. ␥c-specific binding for IL-7 was calculated as the counts/min associated with COS7-mIL-7R␣/ m␥c cells minus the counts/min associated with COS7-mIL-7R␣ cells. ␥c-specific ligand binding to COS7 cells bearing mutant receptors was then expressed as a percentage of the WT control.

RESULTS
Mutagenesis of the m␥c Ectodomain-Previously published models of the ␥c structure (14 -16) predicted that ␥c interacts with its ligands through multiple loop regions pointing toward the elbow-like junction of the N and C domains (Fig. 1A). In particular, the CCЈ1 and EF1 loops of the N domain, residues in the linker region, and the BC2 and FG2 loops of the C domain have been suggested as potential sites for ␥c-cytokine contacts. To experimentally determine m␥c residues involved in ligand binding, we used these models as guides for the site-directed mutagenesis of the m␥c ectodomain. 25 variants of m␥c were generated with mutations affecting a total of 31 residues in the AB1, EF1, BC2, FG2, and linker regions (Fig. 1B). These mutations covered most of the sites that were directly predicted by the models as well as neighboring residues. Initially alanine scanning mutagenesis was performed for single amino acids or pairs of closely spaced residues. Later, additional amino acids within the EF1, BC2, and FG2 loops were mutated, including several substitutions to charged residues. Two cysteines located in the BC2 and FG2 loops were mutated to serine to assess their potential participation in a disulfide bridge. In addition we generated a deletion mutant ⌬(4 -34) lacking residues 4 -34 from the N terminus of m␥c because this nonconserved region was excluded from the homology modeling studies, suggesting that it was dispensable for ligand binding.
mAb Binding to m␥c Mutants-To assess the effects of mutations on the overall conformation and surface expression of m␥c, COS7 cells transfected with hIL-2R␤ and m␥c or mIL-7R␣ and m␥c were tested by FACS analysis for their capacity to bind three mAbs directed to distinct epitopes of the extracytoplasmic region of m␥c (see Ref. 19 and data not shown). The binding of the 4G3 and 3E12 mAbs is known to be conformation-sensitive inasmuch as these reagents detect m␥c only under nonreducing conditions in Western blots (19). Fig. 2 shows representative histograms of selected m␥c mutants whereas Table I represents a summary of this analysis from all the mutants. As expected 10 m␥c mutants (Table I, type A), represented by Y103R (Fig. 2), stained similarly with all three anti-m␥c mAbs and were expressed at levels comparable (50 -100%) to WT m␥c. A second class consisting of seven mutants (type B), represented by D99A/I100A, also stained similarly with all three mAbs, but the overall level of expression was less than 50% of WT m␥c. Finally, the last group of eight mutants (type C), represented by C161S and C210S, expressed varying levels of m␥c and exhibited a selective loss in the binding of the TUGm2 mAb. All these mutations were located between amino acids 158 -162 (BC2 loop) and 206 -211 (FG2 loop) and likely affected residues within or in close proximity of a putative epitope region for TUGm2. Lastly for only one mutation, the ⌬(4 -34)  (16)) illustrating its domain structure and potential ligandinteracting loops is shown. B, sequence of the m␥c extracellular region is shown. Conserved motifs of the type I cytokine receptor superfamily are shown in clear boxes. Predicted ␤-strands are underlined and marked A1 through G1 for the N domain (domain 1) and A2 through G2 for the C domain (domain 2). Shaded boxes indicate mutated residues. Putative ligand-interacting sites predicted by homology modeling are indicated by asterisks.
deletion, a slight selective decrease in 4G3 binding was observed, which may be due to a direct role of this region in the binding of 4G3 or due to conformational effects.
Biochemical Analysis of m␥c-To biochemically investigate the expression of mutant m␥c, detergent extracts of transfected COS7 cells were analyzed by Western blotting using an antiserum specific for the cytoplasmic tail of m␥c. This antiserum detected m␥c on reduced Western blots and hence revealed all m␥c molecules independent of their conformation. With the exception of the ⌬(4 -34) mutant, which was smaller in size (Fig. 3A, lane 3), mature m␥c was detected as an endoglycosidase H-resistant (data not shown) 64 -70-kDa band that is heterogeneous because of variable N-linked glycosylation (22). A second usually fainter endoglycosidase H-sensitive band was also detected with a size that is consistent with unglycosylated or immature precursors located in the endoplasmic reticulum.
It was evident that when compared with WT m␥c, the levels of mature m␥c for some mutants were markedly lower (Fig. 3A, lanes 4, 6, 15, 16, 18, 19, 22, and 23). For several of these mutants (Fig. 3A, lanes 6, 18, 19, 22, and 23) an increase in the intracellular precursor form was also observed. Importantly densitometric analysis of the 64 -70-kDa bands of m␥c from these and other Western blots indicated that there was a direct correlation between the maximum anti-m␥c mAb binding to mutant m␥c during FACS analysis (Table I, underlined values) and levels of the mature form of m␥c (Fig. 3B). Therefore, the reduced staining for m␥c for these mutants truly represents low cell surface expression and is not indicative of impaired mAb binding due to altered conformation of the m␥c.
IL-2 Binding to Heterodimeric IL-2R Containing Mutant m␥c-Although neither hIL-2R␤ nor m␥c can significantly bind IL-2 as individual subunits, the co-expression of these two molecules on the surface of COS7 cells leads to the formation of heterodimeric IL-2R that can readily interact with the ligand (23). Therefore, co-transfection of COS7 with hIL-2R␤ and m␥c was used to study the effects of mutations of m␥c on the binding of IL-2. Data from a typical binding assay are shown in Fig. 4A. As expected, mock transfected COS7 cells and COS7 solely transfected with hIL-2R␤-bound background levels of radioactivity. This binding was nonspecific as verified by the addition of 200ϫ molar excess of cold IL-2 (data not shown). In contrast, COS7 cells co-transfected with hIL-2R␤ and WT m␥c (or most mutant m␥c molecules) readily bound IL-2 in a ␥c-specific manner (Fig. 4A).
Because the cell surface expression of m␥c consistently varied and was lower for some mutants, we determined to what extent the binding of IL-2 was dependent upon the levels of m␥c. For this COS7 cells were co-transfected with a fixed concentration of hIL-2R␤ and variable concentrations of WT m␥c. The cells were then subjected to FACS analysis and IL-2 binding assays. This analysis clearly indicated that IL-2 binding was dependent on the surface levels of m␥c (Fig. 4B). Thus, to evaluate the ability of mutant m␥c to contribute to IL-2 binding, m␥c surface expression also had to be considered. Therefore, we generated a curve (Fig. 4B) that depicts normal IL-2 binding for any given level of WT m␥c surface expression. This was used as a standard curve to compare the cytokine binding of IL-2R containing mutant m␥c to that expected from WT receptors (Fig. 4, D--F). Mutant m␥c molecules whose symbols fall below the standard curve are likely impaired in their ability to enhance IL-2 binding.
Most mutations in the N domain, including the ⌬(4 -34) deletion (Fig. 4D) as well as five mutations in the BC2 loop (Fig.  4E) and four mutations in the FG2 loop (Fig. 4F) had little or no effect on the binding of IL-2 even though there was a variation in m␥c surface expression (Fig. 4C). However, a moderate (2fold) decrease in IL-2 binding was observed when Tyr-103 in the N domain of m␥c (Fig. 4D) was mutated to arginine. This decrease in IL-2 binding was also consistently detected when mutant m␥c Y103R was expressed at varying levels in COS7 cells (Fig. 4C). The Y103A mutation also resulted in a slightly reduced IL-2 binding (Fig. 4D). Because both Y103A and Y103R were expressed normally based on Western blotting (Fig. 3A) and FACS analysis ( Fig. 2; Table I), this region of ␥c likely contributes to binding function.
Strikingly, the mutations C161S and C161A/L162A in the BC2 loop (Fig. 4E) and C210A/G211A, C210S, and G211R in the FG2 loop (Fig. 4F) resulted in a severe loss of IL-2 binding. Although these mutants were poorly expressed, we estimated that IL-2 binding dependent upon m␥c C161S was at least nine times lower than expected for WT m␥c at the same surface level. Similarly, the decrease in IL-2 binding compared with expected values was at least 12-fold for mutant C210A/G211A, 18-fold for mutant C210S, and more than 40-fold for mutant G211R. Moreover, other m␥c mutants with similar expression (e.g. Q163A, Fig. 4E) bound much higher amounts of IL-2 than C161S, C210S, or G211R. Therefore, the integrity of residues Cys-161, Cys-210, and Gly-211 is critical for m␥c participation in IL-2 binding in conjunction with hIL-2R␤.
IL-7 Binding to Heterodimeric Receptors Containing Mutant m␥c-Unlike IL-2R␤, the IL-7R␣ subunit directly binds IL-7 by itself with a dissociation constant (K d ) of approximately 200 pM. ␥c does not directly bind IL-7; however, it cooperates with IL-7R␣ to form a heterodimeric IL-7R, which now binds IL-7 with a 3-fold higher affinity (K d ϭ 80 pM) (7). In an approach similar to that used for IL-2, we investigated the effects of m␥c mutations on IL-7 binding, and a representative IL-7 binding experiment is shown in Fig. 5A. As expected, when compared with mock transfected cells, COS7 only transfected with IL-7R␣ specifically bound IL-7, whereas COS7 co-transfected with IL-7R␣ and WT m␥c or most mutant m␥c chains typically bound higher amounts of ligand. This difference was calculated to represent ␥c-dependent IL-7 binding. The binding to IL-7R␣ or IL-7R␣ and m␥c was also determined to be specific by inhibition using 200ϫ molar excess of cold IL-7 (data not shown).
The expression of m␥c mutants in COS7 cells co-transfected with IL-7R␣ was also variable and typical of what was detected when the same mutants were co-expressed with IL-2R␤ (data not shown). Furthermore, the binding of IL-7 was also dependent on the level of cell surface m␥c expression (Fig. 5B). Therefore, we constructed a standard curve to establish the amount of ␥c-specific IL-7 binding as a function of WT m␥c surface expression (Fig. 5B). Because of the combined errors in the measurement of IL-7 binding to COS7 cells transfected with only mIL-7R␣ and COS7 co-transfected with mIL-7R␣ and m␥c, the equation of the standard curve could not be determined as precisely as with IL-2, especially between 20 and 60% surface m␥c expression.
Based on their effects on IL-7 binding, mutations in the m␥c ectodomain were divided into four groups. Several mutants with normal or slightly reduced surface expression bound IL-7 similarly to WT m␥c (Fig. 5, C-E). This was also true for the ⌬(4 -34) deletion mutant (Fig. 5C). A second group of mutants, Y103R and Y103A (Fig. 5C), expressed normal cell surface levels of m␥c but exhibited reduced IL-7 binding. Thus, Tyr-103 may participate in both IL-2 and IL-7 binding. The third group of mutants, E46A, Y69A/K70A, D99A/I100A, L102R (Fig. 5C), and Y206A (Fig. 5E), showed both decreased cell surface expression of m␥c and lower than expected IL-7 binding. Although these regions may contribute to ␥c function, we are less certain of this conclusion because of the experimental errors in determining ␥c-dependent IL-7 binding at lower surface levels of m␥c. Finally, a fourth group of TABLE I mAb binding to m␥c mutants expressed in COS7 cells Normalized ␥c-specific antibody binding as described under "Experimental Procedures" is shown as percent of the WT control (mean Ϯ S.E., n Ն 3). The surface expression of m␥c in COS7 cells was estimated from the maximum antibody binding observed with an anti-m␥c mAb for each particular mutant (underlined). Mutations selectively inhibiting the binding of either 4G3 or TUGm2 are highlighted in bold. Mutants were classified as type A (expression Ն 50%), B (expression Ͻ50%), or C (Ͼ50% selective loss of TUGm2 binding).
3. Biochemical analysis of m␥c mutants. A, Western blot analysis of selected m␥c variants is shown. Detergent extracts from COS7 cells expressing WT or mutant m␥c (10 6 cells/lane) were subjected to Western blot analysis using rabbit antiserum specific for the cytoplasmic tail of m␥c. B, correlation between flow cytometry and Western blot measurements of m␥c expression is shown. The relative densities of the mature ␥c bands from panel A were plotted as a function of ␥c surface expression estimated from the maximum anti-m␥c mAb binding observed for each mutant (Table I). mutations represented by C161A/L162A and C161S in the BC2 loop (Fig. 5D) and C210A/G211A, C210S, and G211R in the FG2 loop resulted in poor surface ␥c expression and greater than 10-fold reduction in ␥c-dependent IL-7 binding. This decrease in binding was so large that it cannot simply be attributed to assay variability at lower levels of surface expression of m␥c. Therefore, Cys-161, Cys-210, and Gly-211 are critical for the binding of both IL-2 and IL-7.

DISCUSSION
The development of molecular models for the interactions of ␥c with several of its cognate cytokines (14 -16) provides a conceptual framework to investigate the structural basis of the role of ␥c in ligand binding. We have used this information and performed extensive site-directed mutagenesis of the extracellular region of m␥c to identify residues involved in receptorligand interactions in the context of the IL-2R and IL-7R.
All models have excluded the first 34 residues of ␥c because this region shows no homology to other members of the type I cytokine receptor superfamily. Our data directly demonstrate that this portion of ␥c is dispensable for IL-2 and IL-7 binding. All ␥c models also predicted that Tyr-103 is necessary for the binding of IL-2, IL-4, and IL-7, and our data support this hypothesis. Substitutions of Tyr-103 to either alanine or arginine partially decreased IL-2 and IL-7 binding to their receptors, although m␥c surface expression and the binding of three distinct anti-m␥c mAbs remained unaffected. These data suggest that Tyr-103 is not critical for folding, in agreement with its predicted location in the EF1 loop of ␥c. Because Tyr-103 is homologous to known receptor-ligand contact sites in the GHR (Trp-104) (9) and EPOR (Phe-93) (24), we propose that this m␥c residue is directly involved in the binding of IL-2 and IL-7. A second functionally important region of m␥c has been localized to residues Cys-161, Cys-210, and Gly-211, of which mutations resulted in severe inhibition of ␥c-dependent IL-2 and IL-7 binding. These mutations also consistently resulted in low surface expression of m␥c, probably because of erroneous folding and/or decreased protein stability. Nevertheless, our quantitative analysis of ligand binding as a function of m␥c surface expression suggests that it is highly unlikely that the loss of ␥c-specific cytokine binding was simply due to this poor expression. The functional importance of this region is further highlighted by the involvement of residues 158 -162 and 210 -211 in a putative discontinuous epitope for the anti-m␥c mAb TUGm2 that blocks the binding of all ␥c-dependent cytokines (7,20,25). Epitope-mapping studies of the antagonistic PC.B8 mAb that reacts with a discontinuous site on human ␥c are also FIG. 4. IL-2 binding by COS7 cells co-transfected with hIL-2R␤ and mutant m␥c. A, representative IL-2 binding experiment is shown. Radioligand binding assays were performed and ␥c-specific IL-2 binding (light shading) was calculated as described under "Experimental Procedures." The first two bars show mock transfected controls (MOCK) and cells transfected with only hIL-2R␤ (2B). Remaining bars represent COS7 cells expressing hIL-2R␤ and the indicated m␥c. Results are shown as means Ϯ S.E. of triplicate samples. B-F, ␥c-dependent IL-2 binding is plotted as a function of m␥c surface expression. B shows combined data from four independent experiments using COS7 transfectants expressing varying levels of WT m␥c and a fixed amount of hIL-2R␤. These data were fitted by nonlinear regression to the simple binding equation, Y ϭ B max *X/(K d ϩ X), and the result was used as a standard curve (dotted line) to compare m␥c mutated in the N domain (D), the BC2 loop (E), and the FG2 loop (F), or titrated mutant Y103R (C) to WT m␥c expressed at similar levels. In D-F, ␥c-specific IL-2 binding is shown as means Ϯ S.E. (n Ն 3, smallest error bars are not visible). Mean expression was considered as the maximum anti-m␥c mAb binding to each mutant m␥c, taken from Table I. Horizontal error bars have been omitted for clarity. In C, symbols represent binding/ expression data pooled from four independent experiments.

FIG. 5. IL-7 binding to COS7 cells transfected with mIL-7R␣
and mutant m␥c. Data were obtained and analyzed as in Fig. 4, except that mIL-7R␣ was used instead of hIL-2R␤. A, representative IL-7 binding experiment is shown. ␥c-specific IL-7 binding (light shading) was calculated as described under "Experimental Procedures." B-E show ␥c-specific IL-7 binding to receptors containing WT and mutant m␥c. consistent with a role for this region of ␥c in cytokine binding (26). Because Cys-161, Leu-162, Cys-210, and Gly-211 are conserved in all known mammalian ␥c sequences, it is highly unlikely that these residues function directly in antibody specificity but may rather control the local conformation of the epitope. It should also be stressed that receptors carrying the conservative mutations C161S and C210S were just as impaired in their ability to bind IL-2 or IL-7 as those with substitutions to alanine. One plausible interpretation of this result is that Cys-161 and Cys-211 are required for both structural and functional integrity of m␥c via the formation of a disulfide bond connecting the BC2 and FG2 loops of the C domain. In fact, the most recent structural model of ␥c (16) predicted the formation of such an intrachain disulfide bond, based on an improved sequence alignment between human ␥c (h␥c) and GHR. Alternatively, we cannot rule out a direct participation of either of these cysteines in ligand binding because other models predicted interactions between cytokine residues and backbone atoms of Cys-160 in the h␥c (14,15).
To date, more than 90 naturally occurring mutations have been identified in the human ␥c gene of X-SCID patients, of which at least 42 were missense mutations affecting extracellular residues (27). Unfortunately, only a few of these mutants have been characterized in terms of ␥c expression and function. Our study opens the way for a better understanding of this rich reservoir of data. Importantly all human counterparts of the m␥c residues Tyr-103, Cys-161, Cys-210, and Gly-211 have been implicated in X-SCID. Interestingly although our mutations of the m␥c Tyr-103 only partially affected ligand binding, the Y103N substitution in humans resulted in immunodeficiency, suggesting a critical role of this residue in cytokine signal transduction. Numerous X-SCID sites were localized to the BC2 and FG2 loops of h␥c (C160R, L161R, F205C, L209P, C209Y, G210Y, mutations of the WSXWS motif) supporting the functional significance of this region. In addition, several X-SCID mutations in the C domain of h␥c were substitutions to cysteine (R200C, R204C, F205C, and W240C), consistent with a role of disulfide bond formation in the C domain of ␥c.
The functional region composed of the EF1, BC2, and FG2 loops of m␥c shows numerous similarities with other type I cytokine receptors. For example, critical ligand binding sites have been found in the corresponding loops of GHR (28), EPOR (24), IL-2R␤ (29), gp130 (13), granulocyte-CSFR (12), and granulocyte-macrophage CSFR (11). Crystal structures of the GHR (9, 28) and EPOR complexes (30,31) have shown that the core of the ligand binding interface is formed by a cluster of hydrophobic residues surrounded by hydrophilic amino acids. Our results suggest that the binding interface of m␥c may have a similar organization. Importantly Tyr-103 of m␥c is homologous to critical ligand binding sites located in the hydrophobic clusters of GHR (Trp-104) (28) and EPOR (Phe-93) (24,32). Moreover, in the ␥c model that is most consistent with our data (16), the human equivalents of Leu-102, Tyr-103, Gly-211, and the disulfide-linked Cys-161 and Cys-211 form a hydrophobic cluster that may interact with the cytokines. Mutagenesis studies of the GHR (28) and EPOR (24,32) have also shown that although the receptor-ligand interface includes numerous amino acids, only a few "hot spot" residues had major energetic contributions to the receptor-ligand interaction. Therefore, it is not surprising that although 31 residues of the m␥c have been analyzed, only the mutations of four residues had significant effects on ␥c-dependent cytokine binding.
The functional sites identified by us are not necessarily hot spots of ␥c-ligand interaction. Although Tyr-103 was important, none of the mutations in the N domain abrogated ligand binding. On the other hand, the residues in the C domain may have a structural role rather than act as contact residues. Thus, other amino acids with key functions in cytokine binding may exist elsewhere in the ␥c. In this regard it is interesting to note that we have not identified an epitope for the 4G3 mAb that blocks the binding of IL-2 and IL-7 to m␥c (19). We have also not found unique contact residues used by individual cytokines although such cytokine-specific interactions have been suggested, e.g. for the binding of IL-4 (19). Therefore, it will be interesting to determine whether the m␥c residues identified in this study are also required for the binding of IL-4 or other ␥c-dependent cytokines.