Identification of the soluble granulocyte-macrophage colony stimulating factor receptor protein in vivo

Blood samples were obtained with the informed consent of the donors. Utilization of peripheral blood stem cell materials was reviewed and approved by the Ethics Board of the Foothills Medical Center of the University of Calgary.

All cell lines except × 63.Ag8.653 were maintained in RPMI-1640 supplemented with 10% heat-inactivated fetal calf serum (FCS), 1% antibiotic-antimycotic solution, 50 mmol/L glutamine, 1 mmol/L sodium pyruvate, and 1% nonessential amino acids. The human leukemic cell lines U937, THP-1, K562 additionally received 50 μmol β-mercaptoethanol, the GM-CSF dependent human leukemic cell line TF-1 was maintained with rhGM-CSF (R&D Systems), 1 ng/mL. Dihydrofolate reductase^-/-Chinese Hamster Ovary (CHO) cells were supplemented with 10 μg/mL of adenosine, deoxyadenosine, and thymidine. Supernatants from appropriate cells lines were harvested during the exponential phase of cell growth. × 63.Ag8.653 was maintained in DMEM 10% FCS supplemented with 1% antibiotic-antimycotic solution, 50 mmol/L glutamine, 1 mmol/L sodium pyruvate, and 1% nonessential amino acids.

A cDNA corresponding to the extracellular domain of GMRα (eGMRα, amino acids 1-317) was amplified by RT-PCR from mRNA prepared from the GM-CSF dependent human leukemic cell line TF-1 and was subcloned directly into the mammalian expression vector pKCR. eGMRα was also shuttled into a pBluescript vector containing a cDNA encoding human IL-2 (kindly provided by Dr M. Boneville, Nantes) such that eGMRα was upstream of IL-2, in-frame and separated by a sequence encoding the dipeptide ala-gly. The eGMRα/IL-2 sequence was subsequently cloned into pKCR. The pKCR inserts were confirmed by sequencing.

Ten micrograms of the pKCR-eGMRα and pKCR-eGMRα/IL-2 constructs were transfected separately into CHO cells by electroporation at 300 V and 900 μF (Genezapper 450/2500 IBI, Kodak). Stable transfectants were induced by growth of the CHO cells in increasing concentrations of methotrexate (10-50 nmol) and transformants were cloned by limiting dilution. Production of the eGMRα/IL-2 soluble fusion protein by candidate clones was established initially on the basis of the measurement of immunoreactivity in a human IL-2 ELISA (Immunotech Inc, Marseille, France). Production of eGMRα was initially established by Western blot of media conditioned by candidate clones using anti-GMRα antibody SCO4 (see below).

eGMRα/IL-2 was purified in 2 steps. Conditioned media was first loaded onto a carboxymethyltrysacryl gel column and eluted with 50 mmol/L sodium acetate pH 4.5, 0.5 mol/L NaCl. IL-2 immunoreactive fractions were pooled and loaded onto an immunoaffinity column on which anti-IL-2 antibody IL-2.66 (Immunotech Inc, Marseille, France) was grafted onto CNBr-activated sepharose. The column was eluted with 50 mmol/L sodium acetate pH 4.5, 0.5 mol/L NaCl at a flow rate of 30 ml/h and fractions containing the hybrid protein were detected by IL-2 ELISA. Purity was determined to be > 90% by SDS-PAGE and Coomassie Blue staining and Western blot analysis with anti-IL-2 antibody IL-2.66. The fusion protein was quantitated by amino acid analysis.

BALB/C mice (Iffa Credo, Les Oncines, France) were immunized intraperitoneally twice, at 3 weekly intervals, with 5 μg of recombinant eGMRα/IL-2 in complete Freund's adjuvant emulsified in 0.1 mL of sterile PBS. Spleen cells were fused with mouse myeloma cell line × 63.Ag8.653 using polyethylene glycol 1500, and hybridomas were established by conventional HAT selection. Anti-GMRα antibodies were selected by the detection of immunoreactivity against microtiter wells coated with eGMRα/IL-2 and the absence of immunoreactivity to microtiter wells coated with recombinant IL-2. Antibodies were purified with protein A sepharose (Pharmacia, Uppsala, Sweden) and subclasses of antibodies were determined with a mouse monoclonal antibody isotyping kit (Amersham, Les Ullis, France). Two noncross reactive IgG1 anti-GMRα antibodies, SCO4 and SCO6, were identified for further use.

An enzyme-linked immunoassay (ELISA) was developed using a solid phase coated with anti-GMRα monoclonal antibodies SCO4 and biotinylated-SCO6. 96-well plates were coated with SCO4 at 5 μg/mL in PBS after which the wells were blocked with PBS/BSA 3%. SCO6 antibodies were biotinylated with biotin-ε-amino-caproic acid-N-hydroxysuccinimide ester (Boehringer Mannheim, Germany) following manufacturer's instructions. The ELISA procedure was as follows: 50 μL/well of standard or sample were incubated for 2 hours at room temperature on an orbital shaker. The wells were rinsed 3 times with an automatic washer (SLT, Salzburg, Austria) with 300 μL of a 9 g/L NaCl solution containing 0.05% Tween 80 after which 50 μL/well of biotinylated anti-GMR antibody SCO6 and 100 μL of streptavidin-peroxidase were added. The plates were incubated for 30 minutes at room temperature on an orbital shaker, washed 3 times and 100 μL/well of TMB peroxidase substrate was added. The color reaction was allowed to develop in the dark for 20 minutes with agitation. The reaction was then stopped by addition of 50 μL/well of 2N H₂SO₄ and the absorbance was measured at 450 nm with a microplate reader (Molecular Device, UK). The absorbance of the substrate was subtracted from all values. All determinations were performed in duplicate. For quantitation of solGMRα in plasma samples, polyclonal mouse immunoglobulins (Scantibodies, CA) were added to the biotinylated antibody to a final concentration of 50 μg/mL.

Plasma conditioned by human hematopoietic progenitor cells was obtained from peripheral blood stem cell (PBSC) products collected as previously described9 10 from G-CSF-primed stem cell donors undergoing leukapheresis for autologous or allogeneic stem cell transplant. The PBSC product was collected by a 3- to 8-hour leukapheresis procedure, during which the cells accumulated in a small volume of autologous plasma. The stem cell product was centrifuged at low speed and the cells cryopreserved for later transplantation. The remaining supernatant, consisting of 50 to 200 mL of conditioned plasma, was frozen at −80°C for future use.

Neutrophils from healthy donors were purified by dextran sedimentation, followed by hypotonic lysis and Histopaque centrifugation as previously described.11 Except for the dextran sedimentation step, which was performed at room temperature, the cells were kept at 4°C throughout the isolation precedure. Cell preparations contained > 95% neutrophils with > 99% viability using Trypan Blue dye exclusion. After isolation neutrophils were resuspended at a final concentration of 1 × 10⁷cells/mL in PBS.

Supernatants (ranging from 200-1000 mL) conditioned by HL60, U937, THP1, and K562 cell lines, PBSC-Con A eluates or supernatants conditioned by freshly isolated human neutrophils were applied to a GM-CSF ligand affinity column constructed and used as we have previously described.6 Eluted fractions containing solGMRα were pooled, dialyzed against 1% PBS at 4°C for 8 hours, and lyophilized. The samples were then made up to either 500 μL or 1 mL with distilled water and stored at 4°C or −20°C.

Before passage over the GM-CSF ligand affinity column, the frozen plasma from PBSC products was thawed at 4°C, spun at 16 000 rpm to remove the cryoprecipitate, and gravity filtered using Whatman filter paper to remove any particulate matter. The supernatant was then applied to a Con A sepharose column (Pharmacia Biotech, Uppsala, Sweden) at 4°C at a rate of 40 mL/h. The column was washed with 5 column volumes of binding buffer 20 mmol/L Tris-HCl, 0.5 mol/L NaCl, pH 7.4 and bound glycoproteins eluted with 40 mL of 0.3 mol/L methyl αD-glucopyranoside. The eluate was diluted 3-fold and applied to the GM-CSF ligand affinity column.

Purification of eGMRα from media conditioned by eGMRα-transfected CHO cells and of solGMRα from 1 L of human serum was performed using immunoaffinity chromatography. The anti-GMRα monoclonal antibody SCO4 was grafted onto CNBr-activated Sepharose (Pharmacia Biotech, Uppsala, Sweden), following the manufacturer's instructions, at 5 mg of antibody per milliliter of gel. Samples were loaded at a flow rate of 20 mL/h. The column was washed with 20 mmol/L borate pH 8, 0.15 mol/L NaCl containing 0.05 g/L of Tween 80, and eluted with citrate pH 3 at a flow rate of 30 mL/h. The fractions were neutralized immediately with 1 mol/L Tris pH 10.

Gel filtration chromatography was performed with a calibrated Superdex 75 10/30 column and FPLC system (both from Pharmacia Biotech, Uppsala, Sweden). The sample was eluted with PBS at 0.5 mL/min in 1 mL fractions.

Soluble receptor binding assays were performed and analyzed as previously described.6 

Samples were size fractionated under reducing conditions on 10.5% SDS polyacrylamide gels and electrophoretically transferred onto Immobilon-P polyvinylidene fluoride (PVDF) membranes (Millipore, Bedford, MA) and prepared for Western blotting as previously described.12 The blots were incubated in a 1:5000 dilution of the anti-GMRα monoclonal antibody 8G6 at room temperature for 2 hours. After washing 3 times with TBS-Tween (20 mmol/L Tris, pH 7.6, 137 mmol/L NaCl, 0.1% Tween), the blots were incubated with a horseradish peroxidase-labeled rabbit antimouse polyclonal antibody (Amersham Life Sciences, Oakville Ontario, Canada) for 45 minute at room temperature. After washes with TBS-Tween, blots were visualized with enhanced chemiluminescence detection reagents (ECL, Amersham Life Sciences, Oakville Ontario, Canada) and exposed to x-ray film.

We hypothesized that hematopoietic cells, which expressed tmGMRα, would also express solGMRα. To test our hypothesis we collected supernatant conditioned by 3 human leukemic cell lines known to express tmα (HL60, U937, THP1) and 1 cell line that does not express tmα (K562).13-16 In serial experiments 200 to 800 mL of conditioned media from each cell line was first subjected to ligand affinity chromatography to enhance the possibility of identifying solGMRα. Fractions corresponding to the elution pattern of recombinant solGMRα were thereafter pooled and volume reduced by dialysis and lyophylization. The samples were then analyzed for the presence of solGMRα. As shown in panel A of Figure1 the supernatants of each of the tmGMRα-positive cell lines were shown to contain a GM-CSF specific soluble binding moiety by solution phase receptor binding assays, whereas the tmGMRα-negative K562 cell line did not. As well, Western blot analysis of these samples with the use of an anti-GMRα monoclonal antibody (Figure 1, panel B) showed that the supernatant of each of the cell lines displaying soluble GM-CSF binding activity contained a molecule that was recognized by the anti-GMRα antibody and that comigrated with recombinant solGMRα on SDS-PAGE. The K562 cell line on the other hand showed no immunologic evidence for the production of solGMRα.

To further characterize the soluble GM-CSF binding moiety, saturation binding experiments were performed on the ligand affinity purified supernatants from each of the cell lines. Analysis of the binding data (Figure 2) illustrates that the soluble GM-CSF receptor produced by each of the tmGMRα-positive cell lines demonstrates single site binding characteristics and Table1 shows that the affinity is very similar to that which we have previously documented for recombinant solGMRα (Kd = 2-3 nmol).6 However Table 1 also shows that in all the samples examined the molar quantity of soluble receptors in the binding assay was extremely small ranging from 0.1 to 1.0 nmol. Remembering that the initial volume of supernatant of each of the cell lines was reduced approximately 1000-fold and allowing for losses during chromatography, dialysis, and lyophylization, the initial concentration of solGMRα in the conditioned media was likely 100- to 1000-fold less (1.0-10 pmol).

To examine a more physiologic potential source of solGMRα in an environment in which a very large number of human cells had conditioned the supernatant, we examined plasma conditioned by peripheral blood stem cell products collected for autologous and allogeneic stem cell transplants. Five donors had undergone PBSC harvest by leukapheresis, whereas 1 (patient 4) had been harvested by direct operative removal of bone marrow from the posterior superior iliac crests of the pelvis. 1011 to 1012 nucleated hematopoietic cells had conditioned each sample. As can be seen in panel A of Figure3 GM-CSF specific binding was found in the supernatant from all 6 of the samples. Panel B of Figure 3 demonstrates that this corresponded to the presence of a band migrating with recombinant solGMRα and recognized by anti-GMRα monoclonal antibody 8G6 by Western analysis.

To further characterize the soluble receptor produced in PBSC and bone marrow products saturation binding experiments were performed. However, the amount of soluble receptor produced by any 1 product was found to be too small to allow the performance of accurate saturation binding experiments. To circumvent this, the eluates from all 6 samples that remained after initial analysis were pooled, volume reduced, and then subjected to saturation binding experiments. With this manipulation, we were able to demonstrate that the pooled samples contained a soluble GM-CSF binding moiety with single site binding characteristics, Kd = 7.02 ± 3.25 nmol, n = 3 (Figure4). This is again quite close to the dissociation constant we have established for recombinant solGMRα (2-3 nmol).6 However, even in the pooled, volume-reduced sample the soluble receptor concentration only reached 5.2 nmol.

To examine a more homogeneous physiologic cell source than the PBSC products, fresh human neutrophils were obtained from normal volunteers and were purified to > 95% homogeneity. The cell preps were incubated overnight at 37°C in PBS pH 7.4 and the supernatants were collected and processed as described previously. Material conditioned by 1.1 × 10⁹ neutrophils was analyzed. Figure5 demonstrates that the neutrophil preparations produced soluble GM-CSF binding activity (panel A) and a molecule that was recognized by the anti-GMRα antibody 8G6 and that comigrated with recombinant solGMRα on SDS-PAGE (panel B).

We were disturbed by the lack of correlation between the binding data and the band intensity by Western analysis when examining the various samples for the presence of solGMRα (Figures 1 and 3). To clarify the validity of the data we examined the individual fractions eluted from the ligand affinity column for each of the samples described previously using both binding analysis and Western blot. Figure 6 demonstrates that there was a direct relationship between the amount of binding in each fraction and the intensity of the band by Western analysis. The same direct relationship was shown for each of the cell lines although, as expected, in the K562 cell line no binding or immunologic signal was seen in any fraction (data not shown). However, in the PBSC and neutrophil samples in which we suspected the amount of receptor was extremely low in the individual fractions, the degree of binding and the intensity of the immunologic bands were so weak that no reproducible linear relationship could be established.

To simplify the study of solGMRα in vivo a human solGMRα-specific, ELISA was developed. Two noncross reactive anti-GMRα monoclonal antibodies (SCO4, SCO6) were used in a sandwich technique. eGMRα was used as the recombinant standard after stepwise analysis. First, the signal strength of serial 2-fold dilutions of recombinant IL-2 and eGMRα/IL-2 were compared in a commercial IL-2 ELISA kit (Immnunotech Inc, Marseille, France) and the curves were found to be parallel, suggesting the ELISA recognized both molecules equally. This was confirmed when amino acid analysis of eGMRα/IL-2 revealed a very close concordance between the concentration determined by the IL-2 ELISA and by amino acid analysis (data not shown). Similar parallel dilution curves were found when eGMRα/IL-2, immunoaffinity purified eGMRα and human plasma were analyzed in the solGMRα ELISA suggesting that all forms were recognized equally. eGMRα subsequently replaced eGMRα/IL-2 as the standard in the solGMRα ELISA. Sensitivity, defined as the lowest solGMRα concentration significantly different from the zero standard with a probability of 95%, was 5 pmol in serum, plasma, and tissue culture media.

The assay showed no cross-reactivity with TNFα (10 ng/mL), IL-1α (1 ng/mL), IL-1β (1 ng/mL), IL-2 (10 ng/mL), IL-3 (3 ng/mL), IL-4 (1 ng/mL), IL-5 (650 pg/mL), IL-6 (10 ng/mL), IL-10 (1 ng/mL), GM-CSF (10 ng/mL), or EPO (10 ng/mL). There was also no interference by GM-CSF (10 ng/mL) or by IL-3 (3 ng/mL), IL-5 (650 pg/mL) or EPO (10 ng/mL). Intra-assay precision was determined by examining 3 samples 16 times each on a single ELISA plate (Table 2). Inter-assay precision was determined by examining 3 other samples in duplicate in 8 different assays (Table 2).

The sample sources initially examined by receptor binding assay and Western blot were reexamined with the solGMRα ELISA. As shown in Table 3, the ELISA confirmed the presence of solGMRα in media conditioned by tmGMRα-positive human leukemic cell lines, plasma conditioned by PBSC products, and supernatant conditioned by freshly isolated human neutrophils.

Plasma was collected from normal healthy volunteers and analyzed in the solGMRα ELISA to determine whether circulating levels of solGMRα could be detected. The plasma samples were introduced into the ELISA without manipulation. As shown in Figure7, the amount of solGMRα in human plasma follows a normal distribution with a mean concentration of 36 ± 17 pmol (95% CI = 10-85.5 pmol, n = 47, statistical analysis was performed in MedCalc v4.16f.). To confirm the specificity of the solGMRα signal, plasma samples were incubated overnight at 4°C with recombinant human GM-CSF covalently linked to sepharose beads (NHS-Sepharose 4 fast flow, Pharmacia, Uppsala, Sweden). The plasma samples were subsequently centrifuged to remove the GM-CSF-linked beads and the plasma analyzed in the solGMRα ELISA. The GM-CSF-linked beads successfully depleted the plasma samples of the solGMRα signal (data not shown).

To further characterize the nature of the circulating moiety detected by the solGMRα ELISA in human plasma, 1000 mL of human plasma were first subjected to immunoaffinity chromatography with the SCO4 anti-GMRα antibody. Eluted fractions were evaluated by Western analysis again with SCO4 (Figure 8A) and a band was identified that had an identical electrophoretic mobility as recombinant solGMRα. To further characterize this molecule fraction 9 from the immunoaffinity column was volume reduced and subjected to gel filtration chromatography. Collected fractions were analyzed by solGMRα ELISA. Figure 8B demonstrates that solGMRα derived from plasma has a molecular mass of 50 kd as determined by gel filtration. Both the Western analysis and gel filtration gave values for the size of solGMRα in the circulation, which is in reasonable agreement with the size of recombinant solGMRα6 and with the solGMRα species detected by Western blot analysis of material conditioned by human leukemic cell lines, PBSC products, and freshly isolated human neutrophils.

Plasma samples from 5 patients diagnosed with acute myelogenous leukemia (AML) (kindly provided by Dr C. Chabannon, Institut Paoli Calmette, Marseille, France) and 4 patients with multiple myeloma (MM) (kindly provided by Dr E. Tartour, Institut Curie, Paris, France) were analyzed in the solGMRα ELISA. As shown in Table4, circulating levels of solGMRα were elevated in 4 of the 5 cases of AML and in 1 patient with MM.

The existence and cloning of the alternatively spliced mRNA that encodes the soluble isoform of the GM-CSF receptor α-subunit has been documented for some time now by several groups,3,4,7 and there are a number of reports describing the properties of the recombinant solGMRα molecule in vitro.6,7,12 17-19However, there has been little formal evidence that the solGMRα mRNA supports the production of solGMRα protein in vivo and, thus, the physiologic or pathophysiologic relevance of solGMRα has remained in question. In this article, we provide experimental evidence that the solGMRα protein is produced by hematopoietic cells in vivo and that levels of solGMRα are altered in some patients with hematologic malignancies.

Expression of the cell surface GM-CSF receptor is most commonly associated with hematopoietic cells of the myelomonocytic lineage,13,15,20-24 but has also been documented on nonmyeloid hematopoietic cells25 and a variety of nonhematopoietic cells.26-31 The truncated mRNA that encodes solGMRα has been documented in all cells that also bear tmGMRα, which have so far been examined.7,16 One would therefore predict that solGMRα protein would be a product of these cells. The hypothesis that solGMRα is expressed by tmGMRα-positive hematopoietic cells is supported by our findings with human leukemic cell lines. U937, THP-1, and HL60 cells, all of which bear cell surface GM-CSF receptors,13,15 secrete solGMRα, whereas K562 cells, which are tmGMRα-negative,14,16 could not be shown to produce solGMRα (Figures 1, 2 and Tables 1, 3). Plasma conditioned by PBSC products for autologous and allogeneic transplants also contains solGMRα (Figures 3 and 4). The PBSC products are very rich in GM-CSF receptor-bearing cells at many stages of differentiation but are also very heterogeneous in their cellular constituents, so it cannot be said with certainty what the cellular source of the solGMRα is in the PBSC products. The production of solGMRα by preparations of freshly isolated neutrophils (Figure 5) also seems to support the hypothesis that solGMRα is produced by tmGMRα-positive cells. Human neutrophils have a large population of cell surface GM-CSF receptors compared with other physiologic cell sources.22 However, the neutrophil preps, although > 95% pure, do contain a minor population of mononuclear cells that have previously been shown to contain the mRNA for solGMRα.7 Thus, our data, although supportive of the ability of human neutrophils to produce solGMRα, does not rule out that the contaminating peripheral blood mononuclear cells were responsible for some or all of the solGMRα production in the preparations. In any case, the data indicate that solGMRα is produced and secreted by hematopoietic cells and supports the notion that production is limited to those cells that bear membrane anchored GMRα.

The data also reveal that solGMRα is produced in very small amounts. This fact has made its identification in vivo a challenge. Indeed, our early attempts to identify solGMRα in vivo with unmanipulated conditioned media were wholly unsuccessful. Only after enrichment and purification steps were introduced were we able to make progress. Subsequent quantitation of solGMRα by analysis of saturation binding assays reveals solGMRα production levels in the picomolar concentration ranges for all cellular sources examined. The development of the solGMRα ELISA has greatly simplified the identification and quantitation of solGMRα, and there seems to be good agreement between the data derived by analysis of saturation binding assays and by the ELISA. This is especially true for the human leukemic cell lines. The concordance between the binding analysis and the ELISA for the PBSC products is not as close with the ELISA, suggesting a higher production of solGMRα than the binding analysis. However, the ELISA was performed on unmanipulated plasma conditioned by the PBSC products, whereas the binding assays were performed after ConA sepharose chromatography, ligand affinity chromatography, and dialysis of the eluates, followed by lyophylization. The discrepancy between the 2 methods of quantitation is therefore likely explained by solGMRα losses incurred by the multiple manipulations to allow enrichment of solGMRα before the saturation binding assays. We would suggest that the PBSC solGMRα concentrations derived from the ELISA are likely the more accurate values for the true levels of solGMRα in plasma conditioned by the PBSC products.

The availability of the solGMRα ELISA has afforded the opportunity to establish a range of normal values for plasma levels of solGMRα. Our findings indicate that solGMRα circulates in the blood of normal subjects at concentrations of 36 ± 17 pmol with a range of 10 to 85.5 pmol (n = 47, Figure 7). Interpretation of the levels of solGMRα found in plasma conditioned by the PBSC products must keep in mind this range of solGMRα in normal plasma. Seven of the 8 PBSC products examined by ELISA had solGMRα levels within the normal range for plasma. Thus, despite the fact the plasma had been conditioned ex vivo for 3 to 8 hours by between 1010 and 1012hematopoietic cells, there was little or no increase of solGMRα above normally circulating concentrations. This finding brings into question whether this very large number of hematopoietic cells were producing any significant amount of solGMRα during the collection period and begs the more general question of what cells are the source of the solGMRα detected in plasma. It is possible that the source of circulating solGMRα is not hematopoietic cells but other GM-CSF receptor-bearing cells. For instance, vascular endothelial cells bear GM-CSF receptors and respond to GM-CSF.26 The possibility exists that such nonhematopoietic cells are the source of most of the circulating solGMRα, whereas hematopoietic cells produce solGMRα that is meant to be active only in localized microenvironments. Such a juxtacrine model of activity for hematopoietic cells would require the production of only small quantities of solGMRα, in keeping with what we have documented, but the model is entirely speculative. It is also possible that the conditions of PBSC mobilization inhibit the production of solGMRα by the cells that collect during the leukapheresis procedure.

The availability of the solGMRα ELISA has greatly simplified certain avenues of study of solGMRα; however, the performance of ligand affinity and gel filtration chromatography, saturation binding assays, and Western blot analyses has facilitated substantial characterization of the soluble GM-CSF receptor we have identified in vivo. SolGMRα found in vivo binds to and elutes from a GM-CSF ligand affinity column in the same manner as recombinant solGMRα. In addition, immunoaffinity purification of 1 L of human plasma, using an anti-GMRα antibody, leads to the isolation of a molecule that, on Western analysis with the same anti-GMRα antibody, demonstrates a very similar electrophoretic mobility to recombinant solGMRα (Figure8A). On gel filtration, this plasma derived solGMRα also displays the same size characteristics as recombinant solGMRα (Figure 8B). As well, saturation binding analysis demonstrates that naturally occurring and recombinant solGMRα have identical binding affinities for GM-CSF (Figures 2 and 4).

Examination of some of our data suggests that the story may not be completely told. For instance, there is some variation in the electrophoretic mobility of solGMRα from different sources. Comparison of binding data and corresponding Western blots in Figure 1and Figure 3 also shows that there can be discrepancies between the intensity of the solGMRα immunologic signal and the degree of binding when comparing different cellular sources of solGMRα. However, the linear relationship between immunologic signal strength and binding is preserved when the cellular sources are examined individually (Figure6). The reasons for these differences between cellular sources are unclear, but it suggests modification of solGMRα occurs in a cell specific manner and seems even to vary from one individual to another. Previous work has demonstrated glycosylation of solGMRα,6which gives rise to considerable size heterogeneity and which may contribute to differences in immunologic recognition, depending on the antibody in use. It is also possible that at least some of the solGMRα found in vivo arises by proteolytic cleavage of the membrane anchored GMRα rather than by secretion of the soluble product predicted by the alternatively spliced mRNA. This would lead to a very slightly smaller protein product, although not enough to account for all the observed size discrepancies. Previous work has documented that recombinant solGMRα can oligomerize in solution, but these studies have also shown that it is the monomeric form that binds ligand.17 Thus, some of the size discrepancies may be explained by oligomers of solGMRα, but currently available data would suggest that these forms are not responsible for the lack of concordance between band intensity and binding.

Although the experiments in this article provide evidence that solGMRα is a physiologic molecule in vivo, we also provide data that solGMRα plays a role in pathology. Elevated circulating levels of solGMRα were identified in a small series of patients with hematologic malignancies. Of 4 MM patients, 1 had elevated solGMRα levels but the value was only slightly above the upper limit of normal. On the other hand, 4 of the 5 AML patients had elevated levels, and in 3 of the patients the solGMRα value was considerably outside the normal range. No attempt was made in these experiments to document the exact cellular source of solGMRα in the patients, nor can we provide insight into the consequence of such elevated levels of solGMRα on disease progression. It would seem, however, that at least in a subset of patients with myeloid leukemia and perhaps other hematologic malignancies, solGMRα levels could serve as a marker of disease. It remains to be seen whether the molecule is actually contributing to the development or progression of the disease or is a response of nonleukemic cells to the presence of the disease. Further insight will require examination of a much larger and more comprehensive series of patients.

The potential for cytokine receptors to exist in soluble form has been known for some time now. For many of the cytokine receptors, the initial evidence for the production of soluble isoforms rested on the identification of variant mRNA that predicted the production of a soluble receptor product. Direct evidence of the existence of the soluble receptor at the protein level in vivo in humans exists for a smaller subset of the cytokine receptors. This includes the soluble isoforms of the Growth Hormone Receptor,32,33IL-2R,34 IL-6R,35 EPOR,36CNTFR,37 LIFR,38 gp13 0,39 and IL-5R (F. Montero-Julian, manuscript submitted). Our current article provides direct evidence that solGMRα is also produced in vivo in humans and that it is produced by cells of hematopoietic origin. In addition, we provide evidence that the soluble GM-CSF receptor has a role to play in both physiologic and pathophysiologic states. The challenge now becomes to understand the spectrum of physiologic and pathologic conditions in which solGMRα is altered and to gain insight into the mechanisms of the control of production of solGMRα at the protein level.

We wish to recognize the outstanding technical contributions of Carin Pihl and the secretarial support of Valerie Gilbertson. We also thank Dr Steve Robbins, University of Calgary, for supplying some of the cell line supernatants.