The α Subunit of the Granulocyte-Macrophage Colony-stimulating Factor Receptor Interacts with c-Kit and Inhibits c-Kit Signaling*

The cytokine granulocyte-macrophage colony-stimulating factor (GM-CSF) regulates hematopoiesis and the function of mature host defense cells through the GM-CSF receptor (GMR), which is composed of α (αGMR) and β (βGMR) subunits. Stem cell factor is another important hematopoietic cytokine that signals through c-Kit, a receptor tyrosine kinase, and regulates hematopoietic stem cell maintenance and erythroid development. Like other cytokine receptors, GMR and c-Kit are generally deemed as independent adaptor molecules capable of transducing cytokine-specific signals. We found that the αGMR directly interacts with c-Kit and that the interaction is mediated by the cytoplasmic domains. Furthermore, αGMR inhibited c-Kit auto-phosphorylation induced by the ligand stem cell factor. Consistent with the inhibitory effect, the expression of αGMR was suppressed in cells whose viability was dependent on c-Kit signaling. In contrast, the alternatively spliced α2 isoform of the αGMR could not inhibit c-Kit signaling, providing a rationale for the existence of the α2 isoform. Our results suggest that in addition to having the commonly appreciated roles in cytokine signal transduction, the receptors αGMR and c-Kit could interact to coordinate their signal initiation.

amino acids) of ␣GMR is required for GMR signaling (9 -12). We and others have used genetic and biochemical approaches to find proteins interacting with ␣GMR, and several proteins were identified with different functional implications (13)(14)(15). There is a soluble isoform of ␣GMR as well as an ␣2 isoform (␣2GMR) derived from alternative mRNA splicing (16,17). The ␣2GMR has a different cytoplasmic domain but generally exhibits similar properties to ␣GMR in the cooperation with ␤GMR (17). Unlike the ␣GMR, the ␣2GMR is not up-regulated in granulocyte/macrophage (G/M) lineages, and the ␣2GMR appears to be more relevant in undifferentiated hematopoietic progenitor cells (18). Signaling of GM-CSF through ␣2GMR/ ␤GMR leads to more erythroid differentiation than through ␣GMR/␤GMR, suggesting that the receptor isoforms could guide cell differentiation (18).
c-Kit is a transmembrane receptor tyrosine kinase that regulates the proliferation and differentiation of hematopoietic stem cells, erythropoiesis, melanogenesis, and gametogenesis (19 -22). Stem cell factor (SCF, also known as kit ligand or Steel factor) is the ligand for c-Kit. The intracellular region of c-Kit has two kinase domains separated by a kinase insert (23,24). SCF binding to c-Kit induces receptor dimerization and autophosphorylation, leading to activation of downstream signaling pathways (25). In mice, c-Kit is encoded at the white spotting (W) locus, its ligand SCF is encoded at the steel (Sl) locus, and mutations at the W or the Sl locus lead to sterility, pigment cell deficiency, and macrocytic anemia (26).
c-Kit and ␣GMR are well known markers of differentiation, especially in hematopoietic cells. Starting with a high level of c-Kit and negligible amount of ␣GMR, hematopoietic stem or progenitor cells down-regulate c-Kit and up-regulate ␣GMR when they enter into the G/M lineage (22,27). The receptor subunits are generally deemed as passively regulated adaptor molecules, and a direct interaction between ␣GMR and c-Kit has not been reported. Using our preliminary computational method for predicting protein interactions, we identified c-Kit as a possible interacting protein for ␣GMR. Here we describe how the ␣GMR/c-Kit physical interaction regulates signal initiation of the receptors and how ␣2GMR is different from ␣GMR on modulating c-Kit signaling.

EXPERIMENTAL PROCEDURES
Vectors for 293T Cell Expression-Cells (293T) were maintained in high glucose Dulbecco's modified Eagle's medium * This work was supported by grants from the National Institutes of Health (CA30388), the New York State Department of Health, and the Lebensfeld and Schultz Foundations. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. † Deceased August 9, 2004. 1 To whom correspondence may be addressed. E-mail: jchen10021@yahoo.
Immunoprecipitation-The cDNAs encoding ␣GMR (pcDNA4/His-V5c, 5 g) and c-Kit (in pcDNA4/His-V5c, 5 g) were transfected into 293T cells by calcium phosphate precipitation (28). Antibodies M-14 and C-18 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) were used for immunoprecipitation of c-Kit and ␣GMR, respectively, in lysis buffer (modified phosphate-buffered saline with 135 mM K ϩ , 5 mM Na ϩ , and 1.0% Triton X-100, protease, and phosphatase inhibitor cocktails from Sigma). Protein A (Sigma) was used to capture the antibodies. ␣GMR and c-Kit were detected with anti-V5-HRP (Invitrogen). The ␣GMR and kit256 (C terminus part of c-Kit) interaction experiments used ␣GMR in the pMX vector and kit256 in the pcDNA4/His-V5c vector. ␣GMR was immunoprecipitated and detected with C-18 antibody. The kit256 protein was detected with anti-V5-HRP.
Detection of Phosphorylated c-Kit-Cells (293T) were transfected with mouse c-Kit (0.7 g in pcDM8 vector) alone or in combination with ␣GMR (4 g in pMX vector) for 16 h. Cells were harvested and treated with 100 ng/ml mouse SCF (R&D Systems Inc., Minneapolis, MN) for 10 or 20 min at room temperature. The cells were then lysed in lysis buffer (0.1% Triton X-100), and c-Kit was immunoprecipitated with the M-14 antibody. Protein A was used for capturing the antibody. Phosphorylated c-Kit was detected with anti-phosphotyrosine antibody clone 4G10 (Upstate Biotechnology, Inc. Lake Placid, NY), and c-Kit was detected with the M-14 antibody.
Flow Cytometry Analysis of ␣GMR Expression-TF-1 cells or K562 cells were infected with retroviruses encoding the ␣GMR isoforms and mutants. For flow cytometry analysis, cells were suspended in pH 7.4 phosphate-buffered saline with 1% heat-inactivated fetal calf serum and 0.1% sodium azide (Sigma) containing FITC-conjugated anti-␣GMR antibody S-20 (Santa Cruz Biotechnology, Inc.) for 30 min at 4°C. Cells were then washed with the staining buffer and analyzed with FACScalibur (BD Biosciences). The results were analyzed with FlowJo software.

RESULTS
aGMR Interacts with c-Kit-The short intracellular domain of ␣GMR is required for GM-CSF signaling, and the membrane-proximal proline-rich region is especially important (9 -12). An 18-amino acid peptide of ␣GMR (KRFLRIQRLF-PPVPQIKD) from this region (347-364) was used in the application of our computational protein interaction prediction method. 4   found to be a likely interaction partner with the ␣GMR peptide in reverse orientation (Fig. 1A). Since the sequences of human, mouse, and caprine c-Kit have a very high degree of similarity (Fig. 1B), we tested the interaction between human ␣GMR and mouse c-Kit in co-immunoprecipitation experiments. When co-expressed in 293T cells, human ␣GMR was co-immunoprecipitated with mouse c-Kit (Fig. 1C). We also made a construct with 256 amino acids (kit256) from the C terminus of mouse c-Kit that contains the putative ␣GMRinteracting domain and found that kit256 was co-immunoprecipitated with ␣GMR when co-expressed in 293T cells (Fig. 1D). Therefore the ␣GMR and c-Kit appeared to interact through their intracellular domains.
aGMR Inhibits c-Kit Auto-phosphorylation Induced by SCF-The physical interaction between c-Kit and ␣GMR suggested cooperation or cross-inhibition between SCF/c-Kit and GM-CSF/␣␤GMR signaling. Binding of ␣GMR to c-Kit, most likely to the kinase insert domain, implied that ␣GMR could modulate c-Kit kinase activity, and therefore we investigated whether ␣GMR influenced c-Kit kinase activity in 293T cells expressing mouse c-Kit and human ␣GMR. Mouse SCF stimulated time-and dose-dependent phosphorylation of c-Kit in 293T cells transfected only with c-Kit as detected by anti-phosphotyrosine antibody (Fig. 2). When cells were co-transfected with ␣GMR and c-Kit, the phosphorylation of c-Kit induced by SCF was clearly inhibited (Fig. 2). This result suggested an inhibitory function of the ␣GMR/c-Kit interaction.
SCF/c-Kit-maintained Cells Restrict ␣GMR Expression-We asked whether ␣GMR could inhibit c-Kit signaling in the human hematopoietic cell line TF-1 that responds to growth factors such as SCF, GM-CSF, interleukin-3, or erythropoietin (29). SCF stimulation through human c-Kit maintains the viability and proliferation of TF-1 cells, providing a good model of human c-Kit signaling for testing the ␣GMR inhibitory activity. We generated retroviral expression vectors for ␣GMR, ␣2GMR (the ␣GMR isoform with a variant cytoplasmic tail), ␣GMR-375 (the common portion of ␣GMR and ␣2GMR with 25 amino acids deleted from ␣GMR C terminus), and ␣GMR-356 (44 amino acids deleted from ␣GMR C terminus) (Fig. 3A). The four ␣GMR-related proteins have an identical extracellular domain that can be detected by flow cytometry using the same FITCconjugated anti-␣GMR antibody, conveniently allowing comparison of expression levels. The retroviruses produced in 293T cells were tested in the erythroleukemia cell line K562 that proliferates without requirement for growth factor supplements (31). The four ␣GMR-related proteins were expressed by retrovirus infection of K562 cells with variations in expression levels and infection percentages, possibly reflecting different amounts of retroviruses delivered (Fig. 3B).

When a retroviral vector expresses an inhibitor for SCF/c-Kit signaling in TF-1 cells maintained by SCF, the viability and proliferation of the infected cells should be decreased when compared with cells expressing a non-inhibitory protein. The results showed that in TF-1 cells, ␣GMR-356 (the intracellular domain deletion mutant) and
␣2GMR were expressed at high levels, whereas ␣GMR expression level was low as measured by fluorescence intensity (Fig. 3C). ␣GMR-375 was expressed at an intermediate level  ( Fig. 3C), suggesting that the two domains sequentially deleted from the intracellular portion of ␣GMR are both involved in the interaction with c-Kit. We also normalized the percentages of the TF-1 cells expressing the four ␣GMR-related proteins to those of K562 cells (Fig. 3B), in which we used the same doses of retroviruses. The normalized percentages also supported that ␣GMR inhibited c-Kit and ␣GMR-375 partially inhibited c-Kit (Fig. 3C).

Exogenous ␣GMR Expression in TF-1 Cells Inhibits Endogenous SCF/c-Kit
Signaling-We further attempted to achieve the expression of ␣GMR and ␣2GMR in TF-1 cells and test whether exogenously expressed ␣GMR could inhibit endogenous c-Kit signaling. To allow expression of ␣GMR in TF-1 cells, another growth factor, GM-CSF, was included in addition to SCF, relieving the viability dependence on SCF/c-Kit signaling. We noticed under a microscope that the addition of GM-CSF to SCF caused a significant TF-1 cell size increase and that the bigger cells could be driven back to smaller size in culture with SCF alone, quantifiable in the flow cytometry forward scatter/side scatter (FSC/SSC) plot (Fig. 4A). If TF-1 cells express a protein inhibitor of SCF/c-Kit signaling, SCF should be unable to drive enlarged cells to smaller cells. Therefore we used the cell population ratio of the lower left quadrant (LL, smaller cells) to the lower right quadrant (LR, bigger cells) from the FSC/SSC plot of TF-1 cells treated first with GM-CSF and SCF and then with SCF alone to measure the SCF response. A bigger LL/LR ratio would indicate a higher response. We identified cells positive with ␣GMR, ␣GMR-375, ␣GMR-356, and ␣2GMR with an anti-␣GMR antibody in flow cytometry (Fig. 4B) and analyzed their FSC/SSC plots (Fig. 4C). ␣2GMRand ␣GMR-356-positive cells had an LL/LR ratio similar to non-infected cells (low fluorescence population) (Fig. 4, C and D), indicating that ␣2GMR and ␣GMR-356 did not inhibit SCF/c-Kit signaling. ␣GMR-375 partially inhibited SCF response, and ␣GMR had the strongest inhibition (Fig. 4, C and  D). The results demonstrated that the expression of human ␣GMR but not ␣2GMR inhibited human c-Kit signaling and that the inhibition was dependent on the intracellular domain of ␣GMR. Our results supported that the membraneproximal region of the ␣GMR intracellular domain (used in our protein-interaction prediction) was involved in the interaction with c-Kit. It appeared that the membrane-distal region of ␣GMR intracellular domain enhanced the interaction with c-Kit and that the membrane-distal cytoplasmic tail of ␣2GMR prevented the protein-protein interaction with c-Kit.

DISCUSSION
Our present study defines a molecular interaction between the ␣GMR and c-Kit, which is quite unique in that it involves a receptor tyrosine kinase and a non-kinase receptor subunit. The c-Kit kinase insert domain is known to bind phosphatidylinositol 3-kinase (24), suggesting that it is a domain accessible for interaction with other proteins, and therefore it is understandable that the ␣GMR/c-Kit interaction utilizes their intracellular domains.
Our findings suggest that the inhibition of c-Kit signaling by ␣GMR is a result of direct receptor-receptor modulation. Not only could we roughly confirm the predicted interaction domains, we also demonstrated that ␣GMR inhibited c-Kit auto-phosphorylation in 293T cells with protein overexpression. Auto-phosphorylation is an early step in c-Kit signaling pathways, indicating that the inhibition is at the receptor level. If the inhibition occurred indirectly, i.e. through a third party mediator, the would-be mediator also needs to reach overexpression level to overtake c-Kit. It is very unlikely that 293T cells would happen to have such a mediator at that high level induced by ␣GMR; therefore the ␣GMR and c-Kit interaction appears to be direct. Our data do not exclude the possibility that ␣GMR binding to c-Kit affects SCF binding. Direct receptorreceptor trans-modulations are probably not rare as there are other examples including the hepatocyte growth factor receptor Met inhibition of Fas signaling and the 67-kDa laminin receptor inhibition of GM-CSF receptor signaling (15,32).
As a measurement for SCF-induced cellular responses, we utilized an observation that SCF-supplemented TF-1 cells were smaller than GM-CSF-supplemented cells. The reduction of TF-1 cell size with SCF/c-Kit signaling is consistent with previous reports that c-Kit signaling-deficient mice had severe macrocytic anemia, in which the average size of erythrocytes was larger than normal (21,26).
The expressions of ␣GMR and ␣2GMR at the mRNA level were studied during CD34 ϩ hematopoietic progenitor cell differentiation, and it was found that the ratio of ␣2GMR to ␣GMR was about 30 -60-fold higher in undifferentiated hematopoietic cells (c-Kit positive) than in cells of G/M lineages (c-Kit negative) (18). This implies that the biological function of the alternatively spliced ␣2 isoform is to transduce GM-CSF signal in the presence of c-Kit. Both ␣GMR and ␣2GMR can cooperate with ␤GMR to transduce GM-CSF signals, and thus, their differential effects on c-Kit signaling might be an important biological reason for the existence of the receptor isoforms.
As an inhibitor for c-Kit signaling, ␣GMR would be expected to interfere with biological processes in which c-Kit is known to be important, such as erythropoiesis. It was found that expression of human ␣GMR, but not ␤GMR or interleukin-3 receptor ␣ subunit, in avian primary erythroblasts actively inhibited their outgrowth and erythroid differentiation, in which erythropoietin receptor and c-Kit are involved (33). ␣GMR in those erythroblasts appeared to have induced changes reflective of the anemic phenotype in c-Kit-deficient mice, substantiating that ␣GMR is a c-Kit inhibitor. Another finding indicated that the signaling of GM-CSF through ␣2GMR/␤GMR leads to more erythroid differentiation and less G/M differentiation than through ␣GMR/␤GMR in FDCP-Mix cells (18). Our theory could potentially account for the observed differences in that ␣2GMR allows c-Kit signaling and permits both erythroid and G/M differentiations, whereas ␣GMR inhibits c-Kit signaling and permits predominantly G/M differentiation. Using ␣GMR in the G/M lineage and c-Kit in the erythroid lineage, hematopoietic cells could ensure lineage divisions because the ␣GMR and c-Kit signaling pathways are intrinsically non-coexistent. The regulation of hematopoiesis has been extensively studied, and there are other known hematopoietic lineage regulators such as transcription factors PU.1 and CCAAT/ enhancer binding protein ␣ (C/EBP␣) (34,35). ␣GMR and c-Kit are quite special because these receptors were not previously known to coordinate their own signaling patterns.
We attempted to use the retroviral vectors to express ␣GMR and its variants in purified human bone marrow CD34 ϩ cells and investigate whether ␣GMR could influence lineage differentiation in colony formation assays. The CD34 ϩ cells supplemented with SCF, GM-CSF, interleukin-3, and erythropoietin could express ␣2GMR, ␣GMR-375, and ␣GMR-356 but not ␣GMR in 24 h, similar to SCF-supplemented TF-1 cells (data not shown). The incompatibility of ␣GMR expression in CD34 ϩ cells probably reflects the importance of c-Kit signaling for the maintenance of CD34 ϩ cells (22). The suppression of ␣GMR expression in c-Kit signaling-dependent cells suggested that some differentiation markers can be naturally incompatible in hematopoietic stem cells, providing the stem cells a possible maintenance mechanism.
Although the ␣GMR-expressing G/M lineage cells come from the c-Kit-expressing hematopoietic stem or progenitor cells, intermediate cells with both c-Kit and ␣GMR are rare (27), suggesting that the intermediate cells only exist for a short duration during differentiation. Therefore a natural cell system with both ␣GMR and c-Kit was mechanistically not readily available for us to demonstrate the endogenous ␣GMR/c-Kit association. Nevertheless, it can be inferred from our results that overexpression of ␣GMR is not required for inhibiting endogenous c-Kit signaling since SCF-supplemented TF-1 cells could only allow very little ␣GMR expression, suggesting that the amount of ␣GMR needed for inhibiting TF-1 endogenous c-Kit signaling is very low.
In summary, our studies showed that the ␣GMR could interact with c-Kit and inhibit c-Kit signaling, but ␣2GMR could not inhibit c-Kit signaling. The receptors ␣GMR and c-Kit are more than regulated adaptor molecules for transducing cytokine signals; rather, they are signal regulators in their own right.