INTRODUCTION

The hematopoietic cytokine granulocyte-macrophage colony-stimulating factor (GM-CSF)¹ influences the growth, differentiation, and functional activity of neutrophils, macrophages, and eosinophils (1, 2, 3, 4, 5). GM-CSF stimulates these activities through its interaction with a highly specific receptor expressed in low numbers on the surface of responding cells (6, 7).

The receptor for GM-CSF belongs to a cytokine receptor superfamily characterized by conserved extracellular structures (8). In addition to conservation of structure, the cytokine receptor superfamily also exhibits several trends. The cell surface receptors are multisubunit complexes with most being composed of a ligand-specific  subunit and a nonligand binding  subunit that is responsible for providing the receptor complex with signal transduction capabilities. As well, many of the cytokines that have overlapping functions share subunits among their receptor complexes. For example, the receptors for IL-3, IL-5, and GM-CSF each contain a ligand specific  subunit exhibiting low affinity binding; however, all three utilize a common  subunit to convert the receptor to a high affinity binding complex and to initiate signal transduction (reviewed in Ref. 9).

Another common feature of many cytokine receptors is the formation of soluble isoforms (reviewed in Ref. 10). These have been described for many receptors including those that bind granulocyte colony-stimulating factor, GM-CSF, IL-4, IL-5, IL-6, IL-7, ciliary neurotropic factor (CNTF), and erythropoietin. Although the generation of soluble isoforms is common for many cytokine receptors, the mechanism used to generate them varies. They include alternative splicing of RNA transcripts, proteolytic cleavage of extracellular domains, and cleavage of glycosyl-phosphatidyl-inositol membrane anchors. Several of these soluble cytokine receptors have been detected in human body fluids in both physiological and pathological conditions, suggesting that these soluble isoforms have a physiological role. Several functions have been proposed for soluble receptors ranging from ligand protection and transport to conferring new cytokine responsiveness to cells or antagonizing the activities of the cytokines.

The receptor for GM-CSF is a cell surface membrane-spanning heterodimer. The  subunit (tmGMR) is ligand-specific and binds GM-CSF with low affinity (K_d = 2-8 nM) (11). The  subunit (GMR) is incapable of binding GM-CSF on its own, but in conjunction with GMR, GMR forms a high affinity binding complex (K_d = 50-100 pM) (12, 13, 14, 15). Recent studies have shown that the  and  subunits can only be co-immunoprecipitated in the presence of GM-CSF (16). It is likely that in the absence of GM-CSF, GMR and GMR remain as separate but perhaps spatially close molecules on the cell surface and GM-CSF induces GMR and GMR interaction to form a receptor complex. Current evidence suggests that the  subunit can transduce signals; however, controversy remains over the GMR subunit's role in signaling (17, 18, 19, 20, 21, 22).

The  subunit of the GM-CSF receptor exists in both transmembrane (tmGMR) and soluble forms (solGMR). Examination of the genomic structure of the GMR gene reveals that exon 11 encodes the transmembrane domain and alternative splicing of this exon would result in a 97-nucleotide deletion (23); the precise sequence deleted in the solGMR cDNA (24, 25, 26). This deletion also causes a frameshift that results in premature termination of translation and the formation of a unique 16-amino acid carboxyl-terminal tail. The RNA message for solGMR has been demonstrated in unfractionated bone marrow, peripheral blood mononuclear cells, T-lymphocytes, and fibroblasts from the synovial tissue of rheumatoid arthritis patients and unfractionated synovial tissue from osteoarthritis patients (27) as well as several human myeloid leukemic cell lines (28); however, the protein has yet to be detected in vivo.

Several groups have cloned the cDNA for solGMR (24, 25, 26). Recombinant solGMR is a 55-60-kDa glycoprotein that binds to GM-CSF in solution and antagonizes the activity of GM-CSF when added exogenously to assays of GM-CSF biological activity (27, 29). The data suggest a mechanism of antagonism whereby solGMR binds GM-CSF in solution and functions by keeping the ligand away from its cell surface receptors, indicating that exogenous solGMR does not interact with GMR. In this paper, we describe studies directly testing the potential for solGMR to interact with GMR in the presence or the absence of GM-CSF in vitro. We demonstrate that solGMR added exogenously to BHK cells expressing GMR does not form a cell surface complex with GMR; however, co-expression of solGMR and GMR leads to retention of solGMR on the cell surface and the formation of a functional GM-CSF binding complex. In addition, the cell surface expression of solGMR is independent of the presence of GM-CSF.

MATERIALS AND METHODS

The cloning of human tmGMR and solGMR and the establishment of stable BHK cell lines overexpressing tmGMR or solGMR has been previously described (29). The human GMR cDNA (gift of Dr. K. Kaushansky, University of Washington) was cloned into the mammalian expression vector pDx (Zymogenetics Inc., Seattle, WA) to produce pDGMR. Sequencing of this clone revealed the presence of a single nucleotide substitution (A  C) resulting in a substitution of Ala for Glu at amino acid position 9 of the mature GMR peptide (13).

pDGMR was co-transfected into BHK cells with a vector bearing the dihydrofolate reductase gene (pZEM229R, Zymogenetics Inc., confers methotrexate resistance) to establish GMR-expressing lines. In addition, pDGMR was co-transfected with a vector containing the cDNA for tmGMR (pDtmGMR) and pZEM229R or pZEM229RsolGMR. Transfections were performed using the calcium phosphate precipitation method (30). Clonal cell lines expressing GMR or GMR and GMR were established by selection in 500 µM methotrexate (David Bull Laboratories Pty. Ltd., Mulgrave, Victoria, Australia). Colonies were screened for the cell surface expression of GMR by fluorescence-activated cell sorting using a mouse anti-human GMR monoclonal antibody (3D7, gift of Dr. A. Lopez, Adelaide, Australia). BHK cells co-transfected with the cDNA for solGMR and GMR were grown for 2 months before being sorted for GMR expression using flow cytometry. Stable cell lines were maintained in Dulbecco's modified Eagle's medium/Ham's F-12 medium, 10% fetal calf serum, and 1% antibiotic-antimycotic solution (Life Technologies, Inc., Mississauga, ON, Canada).

Cell surface-associated receptor binding assays and soluble receptor binding assays were performed as described previously (29). The soluble receptor binding assay is designed to specifically precipitate complexes of solGMR/GM-CSF in the presence of PEG 6000 in a manner that allows for the receptor complexes to be separated from free components using centrifugation. The specific activity of the ¹²⁵I-labeled human recombinant GM-CSF (DuPont NEN, Mississauga, ON, Canada) was determined using a human GM-CSF enzyme-linked immunosorbant assay and was found to differ from the values provided by the manufacturer. All experiments were analyzed using enzyme-linked immunosorbant assay determined GM-CSF specific activity (422-1294 µCi/µg). The results of hot saturation receptor binding experiments (¹²⁵I-GM-CSF concentration varied, and unlabeled GM-CSF was constant at >50-fold excess) were analyzed using the LIGAND software program (RADLIG v 4.0 for PC, Biosoft Inc., Cambridge, UK) to determine the statistically favored binding model, numbers of receptors, and dissociation constants.

Preparation of solGMR

Soluble GMR was purified to near homogeneity as described previously (29). The control buffer used in all experiments consisted of a late fraction collected from column purification of solGMR and contained no GM-CSF binding activity.

Exogenous solGMR Interaction

Column purified soluble GMR was added to 10⁶ GMR-expressing cells in binding reactions to give a concentration of 2.8 nM solGMR (35,000 soluble receptors available to each cell, determined by Scatchard analysis) in a 25-µl volume. In addition, GMR-expressing cells were co-cultured with an equal number of solGMR-expressing cells and were tested repeatedly over a period of 2 months for the presence of solGMR on the cell surface using flow cytometry and cell receptor binding assays.

After washing with 2 × 5 ml of phosphate-buffered saline (PBS), cells were detached from the tissue culture plates using Puck's solution (5.4 mM KCl, 140 mM NaCl, 4.2 mM NaHCO₃, 5 mM dextrose, 10 mM HEPES, 1 mM EDTA) and rewashed three times with PBS. Cells were incubated with 0.5 µg/ml primary antibody (anti-GMR antibody 3D7 or mouse anti-human GMR monoclonal antibody (CDw116), Pharmingen, San Diego, CA) in 1 ml of PBS for 30 min at 22 °C. After a further three washes in Dulbecco's modified Eagle's medium/Ham's F-12 medium to remove unbound primary antibody, cells were incubated in 1 µg of fluorescein isothiocyanate-conjugated goat anti-mouse secondary antibody (Becton Dickinson, Mississauga, ON, Canada) in a minimal volume for 30 min at 22 °C. After washing three times in PBS to remove unbound secondary antibody, cells were resuspended in PBS containing 1 mM EDTA and were analyzed in a flow cytometer (Becton Dickinson).

RESULTS

Cell Lines Expressing GMR

To examine the interaction between exogenously added solGMR and GMR in the presence of GM-CSF, we developed a BHK cell line that stably expresses GMR. The presence of GMR on the cell surface was confirmed by fluorescence-activated cell sorting analysis using an anti-GMR monoclonal antibody (data not shown). The surface fluorescence level of GMR-expressing BHK cells is approximately 100-fold greater than that of GMR BHK cells incubated with the secondary antibody alone or control BHK cells.

Interaction of Exogenous solGMR with Cells Expressing GMR in the Presence of GM-CSF

To examine for the potential interaction of exogenous solGMR with cell-surface GMR in the presence of GM-CSF, we incubated our GMR-expressing cells with ¹²⁵I-GM-CSF in the presence or the absence of a >50-fold excess of unlabeled GM-CSF in addition to column purified solGMR or control buffer (Fig. 1) and looked for specifically bound ¹²⁵I-GM-CSF (cell-associated counts that could be competed off by unlabeled GM-CSF) on the GMR cells. Untransfected BHK cells showed no specific binding of ¹²⁵I-GM-CSF, and this was not altered by the presence of either control buffer or solGMR. The tmGMR-expressing cell line specifically bound GM-CSF and served as the positive control. The cell associated counts on tmGMR-expressing cells decreased in the presence of exogenous solGMR, confirming that solGMR can compete with tmGMR for the binding of GM-CSF (29). The control buffer had no effect. Our GMR cell line failed to show specific binding of GM-CSF even in the presence of exogenously added solGMR indicating that although exogenous solGMR is capable of binding GM-CSF (as shown by competition with tmGMR for GM-CSF binding), this ligand-soluble receptor complex is unable to interact with GMR on the cell surface.

Interaction of exogenous solGMR with GMR-expressing cells in the presence of GM-CSF.

To further test the ability of solGMR to interact with GMR, we co-cultured equal numbers of GMR- and solGMR-expressing cells and tested for the presence of soluble receptor on the cell surface by cell receptor binding assays and flow cytometry. The co-cultured cells were unable to bind GM-CSF and GMR could not be demonstrated on the cell surface by flow cytometry (data not shown). In preliminary experiments conducted to determine the sensitivity of the binding assay, we could detect as few as 12 ± 2 receptors/cell on the cell surface (data not shown).

Cell Surface Expression of solGMR

Studies done by several groups have shown the presence of solGMR transcripts in GMR-expressing cells (27). To recreate this potential for co-expression in vitro, we co-transfected BHK cells with the cDNAs for both GMR and solGMR. We confirmed the presence of cell surface GMR using flow cytometry and the expression of solGMR by performing soluble receptor binding assays on cell culture supernatants from the solGMR/GMR co-transfected BHK cells (data not shown).

We examined the ability of solGMR/GMR co-transfected cells to bind GM-CSF in cell receptor binding assays (Fig. 2) and found specific cell surface binding of ¹²⁵I-GM-CSF. Specific cell-surface binding was absent in untransfected BHK cells and cells expressing either GMR alone or solGMR alone. Several different solGMR/GMR clones were examined, and all were capable of binding GM-CSF.

Binding of GM-CSF by solGMR or solGMR/GMR transfected cells.

Comparison of Binding Characteristics of tmGMR/GMR and solGMR/GMR

Receptor binding assays were performed with the solGMR/GMR-expressing cells (co-transfected cell line) and the tmGMR/GMR-expressing cells (co-transfected cell line and HL-60 cells) to directly compare the dissociation constants (K_d) and the numbers of receptors (r) for these receptor complexes. The results were analyzed using the LIGAND software program, and representative Scatchard values are shown in Fig. 3. The co-transfected tmGMR/GMR cells exhibit a single class of high affinity binding site (K_d = 64 ± 8.7 pM; r = 529 ± 24; n = 2) similar to HL-60 cells (K_d = 59 ± 10 pM; r = 79 ± 32; n = 3). The affinity of GM-CSF for solGMR/GMR was intermediate (K_d = 331 ± 56 pM; r = 127 ± 55; n = 4) to the affinities of GM-CSF for the low (solGMR, K_d = 3.8 ± 2.5 nM; n = 3) (29) and high (tmGMR/GMR) affinity GM-CSF receptors. Scatchard analysis of solGMR/GMR also revealed a single class of binding sites.

Binding characteristics of GMR-expressing cells.

Ligand-independent Expression of solGMR on the Cell Surface

To confirm the presence of solGMR on the surface of the solGMR/GMR cell line and to determine the ligand dependence of this interaction, we tested for the presence of solGMR in the absence of GM-CSF by flow cytometry using an anti-GMR monoclonal antibody (Fig. 4). The GMR cell line and cells expressing solGMR alone were negative for cell surface GMR expression. In contrast, solGMR/GMR-expressing cells demonstrated an approximately 100-fold increase of cell associated fluorescence intensity, indicating the presence of solGMR on the cell surface. The presence of cell surface solGMR correlated with the expression of GMR and occurred in the absence of GM-CSF. Reverse transcription polymerase chain reaction was used to confirm that only solGMR and not tmGMR was being expressed in these cells (data not shown).

Cell surface expression of solGMR.

DISCUSSION

We have shown that exogenous solGMR in the presence of GM-CSF is incapable of interacting with GMR expressed on the cell surface. In contrast, when solGMR and GMR are co-expressed in BHK cells, we can demonstrate the retention of solGMR on the cell surface in the presence or the absence of GM-CSF, suggesting that GMR anchors solGMR on the cell surface in a ligand-independent manner.

The  subunit of the GM-CSF receptor was first described as a 400-amino acid ligand binding protein expressed on the surface of GM-CSF-sensitive cells (11). Since the initial characterization of the tmGMR cDNA, many groups have described isoforms of this transcript including a second transmembrane form (GMRB) (26), which differs from the original transcript in its cytoplasmic tail, and a soluble form (solGMR), which lacks the transmembrane domain and has a unique 16-amino acid carboxyl-terminal tail (24, 25). The soluble isoform of GMR is likely generated by alternative splicing of GMR RNA (23), a mechanism commonly used for the generation of soluble cytokine receptors (10). This is in contrast to receptors like the IL-1R soluble isoform, which may be generated by proteolysis of the extracellular portion of the mature receptor (31), and the soluble CNTF-R, which is released from the cell surface by cleavage of the glycosyl-phosphatidylinositol linkage by phospholipase C (32).

In vitro studies have shown that soluble cytokine receptors exhibit contrasting functions. Several have antagonist properties (20, 27, 33, 34, 35, 36) as exemplified by the solIL-5 receptor  subunit. This soluble receptor can bind IL-5 in solution and inhibit IL-5-dependent proliferation and differentiation (37). However, solIL-5R does not interact with the cell surface anchored IL-5R  subunit in the presence of IL-5 (33). In contrast, the soluble CNTF receptor exhibits agonist properties by binding CNTF in solution and subsequently interacting with cell surface anchored leukemia inhibitory factor receptor and gp 130 receptor subunits, the other components of the CNTF-R that are incapable of binding CNTF on their own (36). This interaction imparts new CNTF sensitivities to leukemia inhibitory factor receptor and gp130-expressing cells, which were previously unreactive toward CNTF. The soluble IL-6R shows a similar ability in that solIL-6R in the presence of IL-6 can interact with its  receptor component, gp130, and induce an IL-6 response in gp130-expressing cells normally unresponsive to IL-6 (34).

To date, no soluble receptor has been shown to exhibit both antagonist properties and binding capabilities to their  subunit counterparts. Previous studies done by us and others have shown that solGMR, when added exogenously to GM-CSF-dependent bio-assays, antagonizes GM-CSF activity (27, 29). We undertook experiments to characterize the mechanism of this antagonism by solGMR. By establishing a GMR-expressing cell line, we were able to examine the ability of exogenous solGMR to interact with cell surface GMR in the presence of GM-CSF. We failed to see any interaction of exogenous solGMR with cell surface GMR despite creating an assay that was capable of detecting as few as 12 ± 2 receptors/cell while providing approximately 35,000 solGMR for each GMR-expressing cell. In addition, we could not detect cell surface solGMR after co-culturing GMR-expressing cells with cells expressing solGMR. This observation is similar to the characteristics exhibited by solIL-5R (33).

Many cells of hematopoietic origin have been found to express mRNA transcripts for both solGMR and GMR (27, 28). We wanted to reflect this observation in an in vitro system and developed cell lines that co-express solGMR and GMR. Using these cells, we were able to detect expression of solGMR not only in the cell culture supernatant conditioned by solGMR/GMR co-transfected cells but also on the surface of these cells. The cell surface expression of solGMR resulted in functional GM-CSF binding; however, the retention of solGMR on the cell surface was independent of GM-CSF. The solGMR subunit exhibits both antagonistic properties and is capable of binding its  receptor subunit counterpart.

Controversy remains as to whether tmGMR and GMR exist as preformed complexes on the surface of cells. Recent evidence indicates they do not (16), and we cannot co-immunoprecipitate tmGMR with GMR from co-transfected BHK cells unless GM-CSF is present.² This suggests that the binding of GM-CSF induces stabilization or assembly of the high affinity GMR complex. In contrast, the data presented in this paper suggest a direct, GM-CSF-independent interaction of solGMR with GMR seen only when these subunits are co-expressed. Perhaps co-expression of solGMR and GMR induces conformational changes in one or both of the subunits that reflect possible changes induced by GM-CSF binding to tmGMR in the presence of GMR during high affinity complex formation on the cell surface.

Ligand-independent interactions of tmGMR and GMR have been described by Ronco et al. (38), who made a tmGMR (Cys¹³⁶  Ser) mutant that was unable to bind GM-CSF but formed a high affinity GMR complex when co-expressed with GMR. They suggest this observation implies that tmGMR/GMR exist as preformed complexes on the cell surface; however, immunoprecipitation studies in the presence or the absence of GM-CSF were not done to confirm this.

It is possible that the mutation (Ala for Glu) present in our GMR clone is responsible for the cell surface retention of solGMR in the co-transfected cell line. However, we utilized this GMR clone to produce the co-transfected tmGMR/GMR cell line, and Scatchard analysis of ¹²⁵I-GM-CSF receptor binding assays (Fig. 3) indicates that this cell line forms functional high affinity GM-CSF receptor complexes with a binding affinity comparable with HL-60 cells and with values reported widely in current literature (7, 12, 13, 14, 15, 39). We feel this observation suggests that at least with respect to GM-CSF binding, the GMR (A  C) clone behaves similarly to wild-type GMR and that GM-CSF-independent cell surface retention of solGMR in the co-transfected cell lines is not merely a result of the GMR mutation.

Most hematopoietic cells that bind GM-CSF do so with either low and/or high affinity (7, 39, 40), likely reflecting the presence of tmGMR alone and/or tmGMR/GMR complexes. Soluble GMR has yet to be detected in vivo. HL-60 cells, a GM-CSF-sensitive human myeloid leukemic cell line (41), express native high affinity GM-CSF receptors with a dissociation constant (K_d) of approximately 14-60 pM (7, 39). In addition, Heaney et al. (28) have shown that HL-60 cells make the mRNA transcript for solGMR. Interestingly, cross-linking and immunoprecipitation studies of the cell surface GM-CSF receptors on HL-60 cells show the presence of two major bands after gel electrophoresis (7). One corresponds well to the tmGMR subunit; however, the second, smaller band is undefined by the authors and may represent the soluble receptor. A smaller band is also detected by immunoprecipitation and cross-linking studies in the M14 melanoma cell line (42) and in COS cells (40).

We were able to detect solGMR in both cell culture supernatant and on the surface of solGMR/GMR co-transfected cells, suggesting that solGMR is likely in excess and there are no ``free''  subunits left to interact with. This solGMR/GMR cell surface interaction appears to be stable and has GM-CSF binding capabilities of intermediate affinity. Intermediate binding affinity for GM-CSF has also been demonstrated in primary melanoma cells (K_d = 230 pM) and the M14 melanoma cell line (K_d = 720 pM) (42) and in COS cells (K_d = 510 pM) (40). In addition, Cannistra et al. (43) found neutrophils to have a 2-fold lower binding affinity for GM-CSF (K_d = 90 pM) than other high affinity myeloid cells (monocytes, 24 pM; purified normal bone marrow, 39 pM; myeloblasts from patients with AML, 34 pM), whereas others have found neutrophils to have a single class of high affinity GM-CSF receptors (K_d = 17-50 pM) (7, 39). A number of mechanisms can be involved to account for such intermediate affinities; however, it is possible that the intermediate binding affinities observed for some cell types may represent the presence of soluble GMR on the cell surface.

Questions have remained over the physiological significance of solGMR in vivo. Two recent studies have examined the ratios of solGMR to tmGMR transcript levels in vitro in an attempt to elucidate a role for solGMR in vivo (27, 28). Both groups have demonstrated that the ratio of solGMR to tmGMR transcripts varies depending on the cell examined, the state of cellular differentiation and, in some instances, the presence of a disease phenotype, suggesting that the expression of solGMR and tmGMR are independently regulated. These studies support the idea that solGMR plays a significant physiological role in vivo.

Our data represent the first report of a soluble receptor having both antagonistic properties when exogenous to the cell and the ability to interact with its  counterpart when co-expressed. The significance of this observation is unknown. Until studies on the functions of solGMR are performed in vivo, the physiological significance of solGMR remains speculative.