Interacting residues in the extracellular region of the common beta subunit of the human granulocyte-macrophage colony-stimulating factor, interleukin (IL)-3, and IL-5 receptors involved in constitutive activation.

A previous study using random mutagenesis identified an activating mutation in the common β subunit (hβc) of the human granulocyte-macrophage colony-stimulating factor, interleukin-3, and interleukin-5 receptors in which an isoleucine residue (Ile374) in the extracellular region of hβc is replaced by asparagine (Jenkins, B. J., D'Andrea, R., and Gonda, T. J. (1995) EMBO J 14, 4276-4287). To investigate the mechanism by which this mutation (I374N) acts, we employed site-directed mutagenesis to explore predictions based on a structural model of hβc. We focused on possible interactions between Ile374 and other hydrophobic residues in its vicinity and found that replacement of two such residues, Leu356 and Trp358, with asparagine resulted in constitutive activation of hβc. Hydrophilic substitutions at both of these positions and at position 374 resulted in the greatest degree of activation, as measured by the growth rate of factor-independent cells, while hydrophobic substitutions had lesser or no effects. Moreover, these “weak” substitutions appeared to synergize, since factor-independent cells expressing the double mutants I374F/W358F and I374F/L356A showed substantially higher growth rates than the single mutants. Taken together, these results suggest that Ile374 normally interacts with Leu356 and Trp358, and that disruption of these interactions results in a conformational change in hβc that leads to constitutive activity. A model relating this notion to the predicted structure and to ligand- and α subunit-dependent activation of hβc is proposed.

does not bind any cytokine by itself, but is required to confer high-affinity binding when co-expressed with the ␣ subunits (1)(2)(3)(4)(5). The hGMR␣, hIL-3R␣, hIL-5R␣, and h␤c subunits belong to the cytokine receptor family which includes the receptors for many hemopoietic growth factors and other cytokines such as growth hormone (GH), erythropoietin, thrombopoietin (or c-Mpl ligand), IL-2 and IL-6 (reviewed in Ref. 6). Members of this family are characterized by an extracellular cytokine receptor module of about 200 amino acids containing several conserved motifs, including the hallmark WSXWS (Trp-Ser-Xaa-Trp-Ser) motif (7).
The mechanisms by which cytokine binding induces receptors to form active signal-transducing complexes have not been fully elucidated, but there is increasing evidence to suggest that ligand-induced dimerization of receptor subunits is an essential early step in receptor activation. To date, the best characterized active cytokine receptor complex is that of the human GH receptor (hGHR), the tertiary structure of which has been elucidated by x-ray crystallography (8). The active hGHR complex exists as a homodimer in which both hGHR subunits interact with each other and a single GH molecule. Indirect evidence that homodimerization of the receptors for erythropoietin receptor and thrombopoietin (c-Mpl) is similarly essential for signaling has arisen from the isolation of constitutively activated mutants which most likely mimic ligandinduced homodimerization of these receptors (9 -11).
The receptors for a number of other cytokines are more complex and are comprised of two or more distinct subunits. For example, the active IL-6 receptor complex consists of two IL-6 molecules associated with two IL-6-specific ␣ subunits and a dimer of the common signaling subunit, gp130 (12)(13)(14). The complexity of active cytokine receptor structures is further epitomized by the IL-2 receptor, in which IL-2 induces association of a ligand-binding ␣ subunit and two distinct signaling subunits, ␤ and ␥ (15).
With regard to the hGMR, hIL-3R, and hIL-5R, the precise composition or stoichiometry of the active receptor complexes has yet to be determined, although it is becoming clear that hGMR (16) and hIL-3R (17) undergo ligand-induced ␣-␤ heterodimerization. However, several chimeric receptors that contain the cytoplasmic domain of the ␤ subunit and are predicted to undergo ligand-dependent dimerization have been shown to induce cellular proliferation (17)(18)(19)(20). Moreover, we have previously identified an activating mutation in the transmembrane domain of h␤c (V449E) which, by analogy with a similar mutation in the neu oncogene (21)(22)(23), may act by inducing homodimerization of h␤c (24). Taken together, these data imply that dimerization of the intracellular portion of the ␤ subunit is sufficient to initiate cellular proliferation; however, it has not yet been demonstrated that normal GMR, IL-3R, or IL-5R complexes do in fact contain ␤ subunit dimers.
Our previous work has identified two other activating mutations in h␤c. One mutant, I374N (24), substitutes Asn for Ile 374 in the membrane-proximal extracellular domain of h␤c, while FI⌬ contains a small duplication in this same domain (25). Unlike V449E, neither of these extracellular mutants could confer factor independence on the BAF-B03 cell line, implying that they activate h␤c by a different mechanism (24). These findings raise two questions: (i) how do the I374N (and other extracellular) mutation(s) alter h␤c structure and lead to constitutive receptor activation; and (ii) what features of the activated receptor complexes formed by the transmembrane and extracellular mutants are responsible for their different celltype specificities. In the present report, we address the first of these questions. Initially, we examined the specific structural requirements for receptor activation at residue 374 by introducing a range of amino acid substitutions at this position and testing their ability to confer factor independence on a factordependent hemopoietic cell line. We then employed molecular modeling of h␤c to predict which residues might interact with Ile 374 . We show that amino acid substitutions at some of these residues lead to constitutive activation of h␤c, thus implicating them in receptor activation and interaction with Ile 374 . These findings lead us to propose a model in which interactions between residues in the membrane-proximal region of h␤c are involved in both normal and constitutive receptor activation.

EXPERIMENTAL PROCEDURES
Molecular Modeling of h␤c-A model of the fourth domain of h␤c was developed 2 based on the crystal structure coordinates of the human growth hormone-binding protein (8). The sequences of h␤c and domain 2 of the GH-binding protein II were aligned manually and an Indigo computer (Silicon Graphics) was used to run the molecular modeling programs Insight II, Homology, and Discover (Molecular Simulations Inc., San Diego, CA). Coordinates for regions of h␤c thought to be conserved structurally, corresponding to the proposed ␤-strands, were assigned from the homologous backbone coordinates of GH-binding protein II and some side chain coordinates. Additional loops were assigned from coordinates from a library of protein structures. An inspection of the model revealed a well packed hydrophobic core with only moderate steric clashes between the hydrophobic side chains of adjacent strands. Manual and automated methods were used to select appropriate conformations for the hydrophobic side chains of residues proposed to be buried in the core of the h␤c molecule. The model was evaluated for stereochemical parameters using Procheck (26) and is shown in Fig. 1A. The sequences corresponding to the predicted B, C, and F ␤-strands of h␤c are shown in Fig. 1B.
Site-directed Mutagenesis and Construction of Expression Plasmids-The h␤c cDNA used for mutagenesis was that described by Barry et al. (27); amino acids are numbered from the initiating methionine. Site-directed mutagenesis was carried out on single-stranded DNA with mutagenic oligonucleotides using the pAlter-1 system (Promega) in accordance with the manufacturers instructions. All mutations were confirmed by DNA sequencing, following which mutant h␤c cDNAs were subcloned between the BamHI and HindIII restriction sites of the pRUFNeo retroviral expression vector (28).
Retroviral infection was performed using either stably-transfected ⌿2 packaging cells or transiently-transfected BOSC 23 packaging cells as follows: ⌿2 cells were transfected by the calcium phosphate procedure (28) and stable transfectants selected with G418 (400 g/ml). FDC-P1 and BAF-B03 cells were infected as described previously (24) and cells selected in liquid culture medium containing growth factor and G418 at either 1 mg/ml (FDC-P1) or 1.2 mg/ml (BAF-B03). Cells infected with either wild-type hGMR␣ or hIL-3R␣ retrovirus were selected in medium containing puromycin at 2 g/ml.
BOSC 23 cells were transfected essentially as described by Pear et al. (30). Briefly, 1.5 ϫ 10 6 cells were plated onto 60-mm dishes containing 4 ml of medium 18 h prior to transfection. For transfections, 20 g of retroviral DNA was added to each dish containing fresh medium supplemented with 25 M chloroquine. At 7 h post-transfection, the medium was replaced with fresh medium without chloroquine and the cells were incubated for a further 18 h. Infections were performed by cocultivating 3 ϫ 10 5 FDC-P1 or BAF-B03 cells with the BOSC 23 cells for 48 h in the 60-mm dishes containing 4 ml of medium supplemented with 4 g/ml Polybrene. FDC-P1 and BAF-B03 cells were harvested and selected as before. FDC-P1 cell plating experiments in soft agar were performed essentially as described by Johnson (33); cells were washed and mouse GM-CSF (80 units/ml) or G418 (1 mg/ml) was added as required.
Analysis of Receptor Subunit Expression by Flow Cytometry-Expression of h␤c mutants on the cell surface of infected FDC-P1 or BAF-B03 cells was detected by staining with the anti-h␤c monoclonal antibody 1C1 (17) followed by flow cytometry on an Epics-Profile II analyzer (Coulter). High-sensitivity immunofluorescence (34) was performed by incubating cells with primary antibody followed by biotinylated anti-mouse IgG (Vector Laboratories) and streptavidinphycoerythrin (Caltag Laboratories), whereas standard indirect immunofluorescence used fluorescein isothiocyanate-conjugated anti-mouse IgG (Silenus). hGMR␣ and hIL-3R␣ expression was analyzed as above by high-sensitivity immunofluorescence with the monoclonal antibodies 8G6 (35) and 6H6 (36), respectively.
Cell Proliferation Assays-Infected FDC-P1 or BAF-B03 cells were washed twice and triplicate samples of equal cell number (10 3 or 5 ϫ 10 3 ) were cultured in a 96-well microtiter plate with or without appropriate growth factor for 72 h. Cell proliferation was measured by the CellTiter 96 Non-Radioactive Cell Proliferation Assay (Promega).
GM-CSF Binding Assays-Human GM-CSF was radiolabeled by the iodine monochloride method (37) to a specific activity of 8.8 ϫ 10 5 cpm/pmol. Saturation binding assays were performed essentially as described by Woodcock et al. (35). Briefly, duplicate samples of 5 ϫ 10 5 infected BAF-B03 cells were incubated with increasing concentrations of 125 I-labeled GM-CSF ranging from 10 pM to 10 nM for 3 h at 24°C. Cell-bound radioligand was separated from free radioligand by centrifugation through fetal calf serum, and the radioactivity counted on a Cobra 5010 ␥-counter (Packard). Dissociation constants were determined by using the EBDA and LIGAND programs (38) (Biosoft).

Predicted Structure and Molecular
Modeling of h␤c-Determination of the three-dimensional structure of the hGHR (8) together with sequence alignments and structural prediction (7) has lead to the definition of the cytokine receptor module as a fundamental structural unit of the cytokine receptor family. The structure within this module is conserved and consists of two folding subdomains, each containing seven anti-parallel ␤-strands (identified by the letters A-G). These are folded into a "sandwich" comprised of two sheets, with the first containing ␤-strands A, B, and E, and the second containing strands C, D, F, and G (7,8,39). We have used sequence alignment of the extracellular region of h␤c with hGHR and other members of the cytokine receptor family, and comparison with the structure of the hGH-hGHR complex, to derive a molecular model of the membrane-proximal cytokine receptor module of h␤c (8). 2 Fig. 1A shows the predicted structure of the membrane-proximal subdomain (hereafter termed domain 4) of h␤c.
Biological Activity of Ile 374 h␤c Mutants Expressed in FDC-P1 Cells-We have previously reported that substitution of asparagine for isoleucine at residue 374 (I374N) in domain 4 of h␤c (24) resulted in constitutive activation as shown by its ability to confer factor-independent growth on the factordependent hemopoietic cell line, FDC-P1. The model illustrated in Fig. 1A suggests that the Ile 374 residue lies on the C strand and is buried within a hydrophobic region in domain 4 of h␤c. Furthermore, sequence alignment of the extracellular region of h␤c with hGHR and other members of the cytokine receptor superfamily suggest that the hydrophobic nature of the Ile 374 residue in h␤c is highly conserved at the corresponding positions among members of this receptor family (Fig. 1B). Based upon these observations, it is likely that the hydrophilic substitution in I374N would severely distort the conformation of this region.
To further examine the structural requirements for receptor activation, site-directed mutagenesis was used to replace Ile 374 with a series of different amino acids (alanine, phenylalanine, glutamine, and aspartic acid) which were predicted to induce increasing degrees of structural disruption. The Ile 374 h␤c mutant cDNAs were inserted into the pRUFNeo retroviral vector and introduced into FDC-P1 cells, which were then selected for G418 resistance. Flow cytometric analysis of G418-selected FDC-P1 cells using standard indirect immunofluorescence indicated that a significant proportion of infected cells expressed each Ile 374 h␤c mutant on the cell surface ( Fig. 2A).
The ability of each mutant to constitutively generate a proliferative signal was determined by analyzing the infected cells for factor-independent growth. All of the substitutions at position 374, except alanine, were able to confer factor independence on the FDC-P1 cells (Fig. 2B). Although the level of proliferation of cells in the absence of factor was quite low compared to that seen with factor, this was most likely due to the fact that only a subset of each infected cell population expressed h␤c ( Fig. 2A).
To compare the efficiencies with which each mutant could activate h␤c, G418-resistant cells were selected for factor independence in liquid culture. Interestingly, in several independent experiments, the proliferation rate of factor-independent cell populations infected with the I374F mutant was consistently slower than that of cells infected with the I374D, I374Q, and I374N mutants (Fig. 3A), although surface expression was similar in each case (Fig. 3B).
These and subsequent experiments were carried out using polyclonal pools of retrovirally-infected cells. We reasoned that this should give a true indication of the properties of each mutant, as cell-to-cell variation in expression level or responsiveness to mitogenic signals, etc., would be "averaged out" over the cell population. To validate this approach, and to confirm the specific observations on the Ile 374 mutants, flow cytometric analysis was used to identify factor-independent FDC-P1 clones (isolated as colonies in soft agar) that expressed comparable levels of each mutant. Comparison of these factorindependent FDC-P1 clones indicated that clones expressing I374F also proliferated at a significantly lower rate than those expressing the I374D, I374Q, and I374N mutants (data not shown).
Constitutive Activation of Wild-type h␤c upon Substitutions at Leu 356 or Trp 358 Residues-We used a molecular model of domain 4 of h␤c (Fig. 1A) to identify the residues within this domain of h␤c most likely to interact with Ile 374 . Using this approach, four residues, Leu 356 , Trp 358 , Val 412 , and Val 414 were predicted to interact with Ile 374 . The two Val residues are located on the F strand in the same ␤-sheet as Ile 374 (on strand C), whereas the Trp and Leu residues are located on the B strand in the opposing, second ␤-sheet (Fig. 1, A and B). Interestingly, sequence alignment of the extracellular regions of h␤c and other members of the cytokine receptor family indicate that the Trp 358 residue is highly conserved among other cytokine receptors. These alignments also indicate that the hydrophobicity of residues corresponding to the Leu 356 position is conserved in other cytokine receptors, and that the Val 412 and Val 414 residues are part of a conserved motif, "RVRVR" in h␤c, typified by an Arg-Val-Arg sequence (25,40) (Fig. 1B).
We reasoned that if, as predicted, the residues at positions 356, 358, 412, and 414 normally interacted with Ile 374 , then substitutions at some or all of these positions might also disrupt these interactions and result in receptor activation. Since strong activation occurred by replacing the hydrophobic Ile 374 residue with asparagine, we first tested the effect of this substitution at each of the potentially interacting positions to generate the mutants L356N, W358N, V412N, and V414N. Retroviruses encoding these mutants, as well as wild-type h␤c, were used to infect FDC-P1 cells which were then selected for G418 resistance in liquid culture. Initial flow cytometric analyses of the resultant cell populations indicated that cell-surface expression of some h␤c mutants was only detectable by highsensitivity immunofluorescence (data not shown); as a result, subsequent flow cytometric analyses were performed on cells stained by this method (see "Experimental Procedures"). Fig.  4A shows that all of these mutants (as well as I374N, I374F, and wild-type h␤c) were expressed by the G418-resistant cell populations. After several weeks selection for factor independence in liquid culture, only cultures infected with the Leu 356 , Trp 358 , and, as expected, the Ile 374 mutants contained viable, proliferating cells (data not shown). To test the effects of potentially less severe disruptions (akin to I374F), we introduced other, hydrophobic residues at positions 356 and 358 by substituting alanine for Leu 356 (L356A) and phenylalanine for Trp 358 (W358F). Following introduction into FDC-P1 cells, it was seen that these mutants were expressed on the cell surface (Fig. 4A); subsequent culture in the absence of mGM-CSF showed that both induced weak but detectable (see below) factor-independent growth.
To validate these observations, as well as eliminate the possibility that factor independence was a result of secondary activating mutations, colony assays were performed on the infected FDC-P1 cells immediately after infection. As shown in Fig. 4B, and in agreement with the observations from liquid culture, only FDC-P1 cells infected with viruses encoding Leu 356 , Trp 358 , or Ile 374 h␤c mutants produced colonies that grew without mGM-CSF. Although the frequency of factor independence for cells infected with each activated mutant was low, this again is likely to be due to the fact that only a subset of G418-selected cells expressed the h␤c mutants on the cell surface (Fig. 4A). Factor independence was not a result of low-level autocrine growth factor production as conditioned medium from the factor-independent cell pools did not support the growth of uninfected FDC-P1 cells (data not shown). Additionally, the presence of the appropriate full-length h␤c mutant cDNAs in the infected FDC-P1 cells was confirmed by recovery of the entire h␤c fragment by polymerase chain reaction from genomic DNA, followed by restriction enzyme digestions diagnostic of each mutant (data not shown).
The rate of factor-independent proliferation of cells expressing the L356N and W358N mutants was considerably lower than that seen with I374N, as shown by proliferation assays (Fig. 5A). Furthermore, the proliferation rates of factor-independent cell populations infected with the L356A and W358F mutants were severalfold lower even than those of cells infected with the L356N and W358N mutants, although surface expression was slightly higher for the former two mutants (Fig.  5B). Interestingly, the factor-independent colonies (as in Fig.  4B) that arose from FDC-P1 cells infected with the L356A and W358F mutants were significantly smaller than those infected with the other activated mutants (data not shown). Together, these results suggest that the Leu 356 and Trp 358 mutants induced constitutive activation less efficiently than the Ile 374 mutants and that the asparagine substitutions at these positions lead to higher activity than the more hydrophobic (alanine and phenylalanine) substitutions.

Substitutions at Positions 356 or 358 Can Synergistically Enhance Activation of the I374F Mutant-An extension of the notion that disruption of interactions between Ile 374 and
Leu 356 /Trp 358 leads to constitutive activation is that weakly activating mutations at both of the interacting positions might synergize in enhancing receptor activation. We therefore constructed four double mutants by combining I374F with L356A, W358F, and also with glycine substitutions for each of the two valine residues at positions 412 and 414. The two latter double mutants (I374F/V412G and I374F/V414G) were constructed to provide negative controls for synergy. Glycine was chosen as a smaller, non-polar residue to replace valine, in a similar vein to replacing leucine with alanine and tryptophan with phenylalanine in L356A and W358F, respectively; it was unlikely that these substitutions would be activating as even the disruptive asparagine substitutions for Val 412 or Val 414 did not result in activation (Fig. 4B). All four double mutants conferred factor independence upon FDC-P1 cells; however, the proliferation rates of factor-independent cells infected with the I374F/L356A and I374F/W358F mutants were consistently severalfold higher than those infected with the I374F/V412G, I374F/ V414G, or I374F mutants (Fig. 6A). Indeed, the proliferation rates seen with the I374F/L356A and I374F/W358F mutants were similar to that of the strongly-activated I374N mutant. In contrast, the proliferation rates of cells expressing the I374F/ V412G and I374F/V414G mutants barely differed from that of cells expressing the "parental" I374F single mutant. These differences in growth rates could not be attributed to corresponding differences in the level of cell-surface expression of the various mutants (Fig. 6B). Thus these data, and the data of Figs. 4 and 5, are consistent with the notion that activation of h␤c by the Ile 374 , Leu 356 , and Trp 358 mutants is due to disruption of the interactions between Ile 374 and the latter two residues.
Biological Activity of L356N and W358N Mutants in BAF-B03 Cells-We have previously shown that the I374N mutant, while constitutively activated in FDC-P1 cells, is unable to confer factor independence on mouse IL-3-dependent BAF-B03 cells (24). In contrast, another previously identified h␤c mutant, V449E (24), is constitutively activated in both cell types. We reasoned that if the L356N and W358N mutants are activated in FDC-P1 cells by the same mechanism as I374N, then these mutants would also be unable to confer factor independence on BAF-B03 cells. Retroviruses encoding the wild-type and mutant forms of h␤c were therefore used to infect BAF-B03 cells. As shown in Fig. 7A, each mutant was expressed on the cell surface of infected cells. However, proliferation assays  Fig. 2A. For comparison, analyses of uninfected cells and cells infected with wild-type h␤c are also shown. B, frequency of factor independence of FDC-P1 cells infected with h␤c mutants. Cells were washed and plated in agar-containing medium with or without mouse GM-CSF immediately after co-cultivation with transiently transfected BOSC 23 cells. Data are presented as the average number of colonies present on duplicate agar plates seeded with 500 or 5000 cells. Percentage of factor independence is calculated as the percentage of infected, i.e. G418resistant (as determined from plates seeded with 500 cells), colonies that grew on plates, seeded with 5000 cells, in the absence of mGM-CSF.

FIG. 5. Proliferation of factor-independent FDC-P1 cells infected with activated h␤c mutants.
A, proliferation assay of FDC-P1 cells, infected with the indicated h␤c mutants, which had been selected prior to assay for growth in the absence of factor. Also shown are uninfected cells that were washed and assayed in medium without mouse GM-CSF. The inset shows an enlargement of the proliferation profiles of uninfected cells and cells expressing the L356A and W358F mutants. Procedures and axes are as in Fig. 3A. B, flow cytometric analysis of activated h␤c mutant expression on the factor-independent FDC-P1 cells depicted in A. Procedures, nomenclature, and axes are as in Fig. 4A. showed that none of these mutants were able to confer factor independence on BAF-B03 cells in liquid culture (Fig. 7B).
Although the I374N mutant is not constitutively activated in BAF-B03 cells, it is still capable of forming a high-affinity receptor and delivering a proliferative signal, in the presence of human GM-CSF, when co-expressed with the hGMR␣ subunit (24). 3 We therefore examined the ability of the L356N and W358N mutants to behave as wild-type ␤ subunits by superinfecting BAF-B03 cells expressing these mutants with a retrovirus encoding the hGMR␣ subunit. Flow cytometric analysis indicated that the hGMR␣ subunit was efficiently co-expressed with the wild-type and mutant ␤ subunits on the surface of infected cells (Fig. 7A). However, only cells expressing the wild-type, and as expected I374N mutant ␤ subunits were able to proliferate in 0.1 ng/ml hGM-CSF, as shown by proliferation assays (Fig. 7B), and prolonged monitoring of liquid cultures in the presence of either 0.1 or 1 ng/ml hGM-CSF.
To determine whether this lack of proliferation resulted from a loss of high-affinity binding associated with the L356N and W358N mutants, saturation binding studies were performed on cells with 125 I-labeled hGM-CSF expressing hGMR␣ alone, and hGMR␣ with either the I374N or W358N mutant. As shown in Fig. 7C, cells co-expressing hGMR␣ and I374N exhibited both low-affinity (K d ϭ 2 nM) and high-affinity (K d ϭ 78 pM) binding sites. These binding affinities are consistent with those previously determined for cells co-expressing wild-type hGMR␣ and ␤ subunits (2,35). 4 In contrast, only low-affinity binding sites were detected on cells co-expressing hGMR␣ and W358N (K d ϭ 2.1 nM), or as expected, cells expressing hGMR␣ alone (K d ϭ 1.6 nM). We have also examined the ability of BAF-B03 cells coexpressing the L356N or W358N mutants with the hIL-3R␣ subunit to proliferate in response to hIL-3. Although flow cytometric analysis indicated that the hIL-3R␣ subunit was efficiently cell-surface expressed with wild-type and mutant ␤ subunits, only cells expressing wild-type or I374N mutant ␤ subunits proliferated in response to 1 ng/ml hIL-3 (data not shown).

DISCUSSION
We have previously reported that h␤c can be rendered constitutively active by a point mutation in the extracellular region which replaces a conserved isoleucine residue at position 374 with asparagine (I374N) (24). As one way of exploring the mechanism by which this mutation (and possibly other mutations in domain 4), acts, we have utilized a molecular model of part of the extracellular portion of h␤c to design further mutants. In particular, we have focused on possible interactions between Ile 374 and other, neighboring residues in the predicted h␤c structure.
One of the key observations in this work was that replacement by asparagine of Trp 358 or Leu 356 , which are predicted to participate in van der Waals interactions with Ile 374 , also resulted in activation. Potentially less disruptive mutants, in which these residues were replaced with phenylalanine or alanine (in W358F and L356A, respectively), were very weakly activating by themselves. However, we found that these relatively mild changes greatly enhanced factor-independent proliferation when combined with a relatively weak mutation (I374F) at position 374. The results support the prediction that Ile 374 interacts with Leu 356 and Trp 358 , and lead us to suggest that (i) these interactions are normally involved in maintaining the conformation of domain 4, and (ii) that disruption of these interactions leads to a conformational change which results in receptor activation. While the implications of such a model are discussed further below, we will consider several other observations that support this interpretation of our results. First, substitutions at position 374 other than the original asparagine resulted in activation, with those expected to be most disruptive, i.e. other hydrophilic residues, resulting in maximal activation as judged by the growth rates of factor-independent cells. Second, a similar pattern holds for substitutions at positions 356 and 358 in that the asparagine substitutions induced far greater factor-independent growth than the alanine (L356A) or phenylalanine (W358F) substitutions. Third, we note that the interacting residues Ile 374 and Trp 358 /Leu 356 are predicted to lie on ␤-strands C and B, respectively, and so we could generalize that other interactions between these two strands may also be important in maintaining the normal structure of domain 4. Indeed, data from random mutagenesis have shown 5 that Tyr 376 in strand C is also a target for activating mutations. Finally, the fact that like I374N, neither the W358N nor the L356N mutant could confer factor independence on BAF-B03 cells is consistent with a common mode of action.
One rather unexpected result of these studies was that neither the W358N nor the L356N mutant could form a highaffinity receptor on, or elicit hGM-CSF-dependent proliferation of, BAF-B03 cells expressing the hGMR␣ subunit; this is in contrast to I374N which exhibits wild-type function under these conditions. Two possible explanations are that the W358N and L356N mutations prevent interaction with either GM-CSF itself or with the hGMR␣ subunit. The first of these is consistent with the finding of Woodcock et al. (35) that residues adjacent to ␤-strand B, in the B-C loop, were necessary for high-affinity hGM-CSF binding. Similarly, the inability of the W358N and L356N mutants to allow proliferation in response to hIL-3 in the presence of the hIL-3R␣ subunit is consistent with the recent finding that other mutations in the B-C loop interfere with IL-3 binding. 6 However, it is equally possible that these mutations prevent functional interaction with receptor ␣ subunits; this would be consistent with the prediction that contacts between the ␣ and ␤ subunits involve the adjacent A-B loop (8). 2 Ideally, we would like to integrate our observations into a model that could explain how the activating mutations achieve their effect and how this mechanism relates to normal receptor function. Unfortunately, it is not yet known precisely how the GMR is triggered by ligand binding, nor is the stoichiometry of the active receptor complex known (other than that it contains at least one ␣ subunit and one ␤ subunit). It is highly likely, however, that the active normal GMR complex contains more than one signaling subunit, as do the active forms of all other known cytokine receptors; these could be either two ␤ subunits or, as we have suggested elsewhere (24) a ␤ subunit plus a putative heterologous subunit ("␥"). It is therefore likely that one ␤ subunit would both bind ligand in conjunction with an ␣ subunit and also interact with a second ␤ subunit or a "␥ subunit." As mentioned previously, structural modeling of the ␣-␤ligand complex indicates that domain 4 of h␤c can be viewed as two ␤-sheets, one comprised of strands A, B, and E (␤-sheet 1), and the second comprised of strands D, C, F, and G (␤-sheet 2) (8,39). This model, which is supported by extensive studies on the interactions between GM-CSF (and IL-3) and both the ␣ and ␤ subunits (e.g. Ref. 41), predicts that ␤-strand E and the A-B loop in ␤-sheet 1 contact the ␣ subunit (see Figs. 1 and 8B). Thus, we would predict that interactions with a second signaling subunit would take place via the opposite "side" of domain 4, i.e. ␤-sheet 2 (see Fig. 8B). In view of this and the results presented in this report, we propose the following model, illustrated in Fig. 8, for the role of domain 4 in the activation of h␤c. In both inactive and active forms of the wild-type receptor, ␤-sheets 1 and 2, and specifically strands B and C interact via contacts including those between Ile 374 and Trp 358 /Leu 356 (Fig.   5 B. J. Jenkins and T. J. Gonda, unpublished observations. 6 J. Woodcock and A. F. Lopez, unpublished observations.

FIG. 8. Model for the involvement of interactions between ␤-strands B and C of domain 4 in receptor activation.
A, in the inactive, i.e. uncomplexed form, interactions (double arrow) between the two ␤-sheets, comprising ␤-strands A, B, and E (␤-sheet 1) and ␤-strands C, D, F, and G (␤-sheet 2), respectively, stabilize the inactive conformation of domain 4. B, interaction with ␣ subunit plus ligand induces a conformational change in ␤-sheet 1 that is transmitted via the B-C interaction to generate a conformational change in ␤-sheet 2; the altered conformations are represented by increased curvature (compared to A). The altered conformation of ␤-sheet 2 results in association with a second signaling subunit (either another ␤ subunit, ␤Ј, or a possible heterologous signaling subunit ("␥"), and thus triggers receptor signaling. Note that for the sake of clarity, the ligand itself is not depicted. C and D, activating mutations in ␤-strands B or C (depicted by asterisks) disrupt interactions between the two ␤-sheets and result in sheet 2 assuming an activated conformation, which in turn allows interaction with the second signaling subunit (as in part B). A and B). Association with the ␣ subunit plus ligand induces a conformational change in ␤-sheet 1 or a rearrangement of the interface between the two sheets, and this is transmitted via the B-C interaction to the second ␤-sheet (Fig. 8B). The ensuing conformational change (to which contacts between ligand and h␤c could also contribute) then promotes interaction of residues in this ␤-sheet with the second signaling subunit, leading to dimerization and triggering of intracellular signaling pathways. In the case of the activating mutations affecting Ile 374 , Trp 358 , and Leu 356 , disruption of the B-C interaction would lead to the second ␤-sheet assuming an activated conformation similar to that seen after ␣ subunit/ligand binding in the normal receptor (Figs. 8, C and D). Validation or rejection of this model will ultimately require definition of the subunit composition of both wild-type and mutant GMR/IL-3R/IL-5R complexes, and identification of all surfaces participating in intersubunit interactions.

8,
Finally, we note that the three interacting residues studied in this report, Ile 374 , Leu 356 , and Trp 358 , are highly conserved within the cytokine receptor family (e.g. see Fig. 1B). Thus, the homolgous residues in other cytokine receptors may be targets for activating mutations and, furthermore, the model proposed here may also be applicable to other receptors.