Contribution of the Membrane-distal Tyrosine in Intracellular Signaling by the Granulocyte Colony-stimulating Factor Receptor*

We have evaluated the contribution of intracellular tyrosine residues of the granulocyte colony-stimulating factor receptor (GCSF-R) to its signaling and cellular outcomes. We began with stable BaF3 cell lines overexpressing wild-type or mutant GCSF-Rs. When all four intracellular tyrosines of the GCSF-R were replaced with phenylalanine (FFFF GCSF-R), cell proliferation and survival were compromised. Replacement of only the membrane-distal tyrosine (YYYF GCSF-R) also showed reduced survival following a GCSF withdrawal/replacement protocol, suggesting a role for this tyrosine. Proliferation by FFFY GCSF-R cells was attenuated by ∼70%. In evaluating the biochemical steps involved in signaling, we then showed that the membrane-distal tyrosine was necessary and sufficient for c-Jun N-terminal kinase (JNK) activation. With the use of a cell-permeable JNK-inhibitory peptide, JNK was implicated in the proliferation of the FFFY GCSF-R mutant. To further define the events linking the membranedistal tyrosine and JNK activation, the Src homology 2 domains of Shc, Grb2, and 3BP2 were shown to bind the full-length GCSF-R and a phosphopeptide encompassing the membrane-distal tyrosine. When binding to variant phosphopeptides based on this membrane-distal tyrosine was tested, altering the amino acids immediately following the phosphotyrosine could selectively abolish the interaction with Shc or Grb2, or the binding to both Grb2 and 3BP2. When these changes were introduced into the full-length GCSF-R and new cell lines created, only the mutant that did not interact with Grb2 and 3BP2 did not activate JNK. Our results suggest that direct binding of Shc by the GCSF-R is not essential for JNK activation.

Granulocyte colony-stimulating factor (G-CSF) 1 is a cytokine regulating the survival, proliferation, differentiation, and functional maturation of cells committed to the neutrophil-granulocyte lineage (1). To understand these important physiological activities, the intracellular signaling events initiated by the G-CSF receptor (GCSF-R) have been extensively evaluated. The GCSF-R has three domains: the extracellular ligand-binding domain, the transmembrane domain, and the intracellular signal transducing domain, which lacks protein kinase activity but contains three subdomains (box 1, 2, and 3), as well as four tyrosine (Y) residues. In the murine GCSF-R these are Tyr-703, Tyr-728, Tyr-743, and Tyr-763 (2). Following G-CSF binding, the GCSF-R rapidly homodimerizes and is phosphorylated by tyrosine kinases such as the Janus tyrosine kinases (3). The activation of several mitogenic signaling pathways (including Janus tyrosine kinase-STAT and mitogen-activated protein kinase (MAPK) pathways) then follows (4 -6).
Studies utilizing truncated GCSF-Rs have further defined functional regions within the signal transducing domain. Specifically, the region including box 1 and 2 is critical for mitogenic signaling in BaF3 and FDC-P1 cells (7,8), whereas the region containing box 3, Tyr-728, Tyr-743, and Tyr-763 is essential for differentiation of L-GM, FDC-P1, and 32D myeloid cells (7,9). However, there are also conflicting results obtained with this approach (7,10), and so the role of the individual Tyr residues has also been assessed more specifically by mutagenesis within the context of the intact GCSF-R. In this way, each Tyr has been replaced alone or in combination with phenylalanine (Phe) (11)(12)(13)(14). Upon phosphorylation, each Tyr of the receptor would be expected to interact directly with Src homology 2 (SH2) domain-containing proteins, thereby bringing signaling proteins together (15,16). The binding specificity of each SH2 domain is dictated by the three residues immediately C-terminal to the phosphotyrosine (17,18). This explains how the disruption of a single Tyr could specifically affect a defined signal transduction event. Therefore, the membrane-proximal tyrosine (Tyr-703), which is a predicted STAT3 binding site (YXXQ) in the GCSF-R can be disrupted. This Tyr-703 3 Phe mutant displayed reduced ability to activate STAT3 in M1 myeloid leukemic cells (19). Furthermore, disruption of the membrane-distal tyrosine (a Tyr-763 3 Phe mutation) completely abrogated G-CSF-stimulated c-Jun N-terminal kinase (JNK) MAPK activation in BaF3 pro-B cells (6).
In this study, we found signaling initiated by the membranedistal tyrosine of the GCSF-R contributed to cell proliferation and survival, as well as JNK activation in response to G-CSF. Utilization of a cell-permeable peptide inhibitor of JNK showed that JNK mediated the proliferative signal initiated from the membrane-distal tyrosine. Binding partners for the membrane-distal tyrosine were then investigated to evaluate how cytokine receptors initiate JNK pathway activation. Previous data predicting SH2 domain binding partners (18,20) could not unambiguously identify a direct binding partner for GCSF-R Tyr-763, although we confirmed that Shc interacted with the membrane-distal tyrosine of the full-length GCSF-R. We therefore utilized peptides based on the sequence surrounding this phosphorylated membrane-distal tyrosine of the GCSF-R to biochemically identify candidate direct binding proteins. Using BIAcore analyses we identified three adaptor proteins (Shc, Grb2, and 3BP2) as high affinity partners. This analysis was extended to evaluate how alteration of the amino acids immediately following the membrane-distal tyrosine might confer signaling specificity, with these mutations then introduced into the context of the full-length GCSF-R. This analysis revealed that the direct binding of Shc or Grb2 is not required for JNK activation by the GCSF-R.
GCSF-R Mutagenesis and Generation of Stable BaF3 Cell Lines-GCSF-R mutants in which tyrosines were replaced with phenylalanine were mutant human GCSF-Rs. The membrane-distal tyrosine was replaced with phenylalanine (YYYF), only the membrane-distal tyrosine retained (FFFY), or all tyrosine residues replaced (FFFF). The residues following the membrane-distal tyrosine were also mutated within the murine GCSF-R to produce the YEAI, YVNV, and YAAA mutants. The residues C-terminal to the membrane-distal tyrosine were similarly mutated within the human GCSF-R to create the YENQ GCSF-R mutant. A wild-type murine or human GCSF-R cell line was produced with each set of mutants and used as a control. The FFFY, YYYF, and FFFF GCSF-R cell lines were produced by the method described (11), except for co-transfection with the pgKNeo-pA plasmid encoding neomycin resistance. The YEAI, YVNV, and YAAA mutant GCSF-R cell lines were produced by the method described (6). The YENQ mutant was produced by mutation of the full-length GCSF-R cDNA (21) in Bluescript SKϩ using the QuikChange TM site-directed mutagenesis kit (Stratagene). The mutated GCSF-R was then cloned into the XbaI site of the mammalian expression vector pEF-BOS (22) and correct orientation and presence of mutation confirmed by Dye Terminator sequencing (ABI PRISM TM ). Parental BaF3 cells were electroporated at 230 V, 960 microfarads for 1 s with pEF-BOS plasmids containing wild-type or mutated GCSF-Rs and ScaI-linearized pcDNA3.1 containing the neomycin resistance gene (Invitrogen). Following selection in neomycin (1.2 mg/ml), cells stably expressing the GCSF-R were stained using 10 g/ml anti-human GCSF-R antibody LMM741, and PE-goat anti-mouse IgG 1 (␥1) conjugate, and sorted by flow cytometry.
Assessment of BaF3 Cell Survival and Proliferation-The contribution of the membrane-distal tyrosine to proliferation and survival signaling was determined following the creation of stable BaF3 cell lines expressing the wild-type GCSF-R or the Tyr 3 Phe GCSF-R mutants YYYF, FFFY, and FFFF. For all cell lines, expression of comparable levels of GCSF-R was confirmed by flow cytometry using the antihuman GCSF-R antibody LMM741.
For the evaluation of the contribution of the membrane-distal tyrosine of the GCSF-R to survival signaling, cells were deprived of G-CSF for between 2 and 24 h, then G-CSF returned to media for the remaining 24-h period. Viability at 24 h was determined by staining with annexin V-FITC and propidium iodide (PI) (23). Viable cells were those that excluded both reagents.
For the evaluation of the contribution of the membrane-distal tyrosine of the GCSF-R to proliferative signaling, cells were deprived of IL-3 for 2 h and then stimulated with G-CSF (50 ng/ml) or IL-3 (20 ng/ml) for 6 days. Viable cells only (discriminated by exclusion of trypan blue) were counted using a hemocytometer. Results shown are from one clone expressing each receptor mutant; similar results were also obtained in two other FFFF clones and one other clone for the wild-type, YYYF, and FFFY GCSF-R. 2 Where indicated, cells were cultured in the presence of G-CSF with 10 M FITC-labeled cell-permeable inhibitor of JNK (FITC-GRKKRRQRRRPPRPKRPTTLNLF; Tat-TIJIP; Auspep) based on the human immunodeficiency virus Tat protein transduction domain (24) and a peptide JNK inhibitor (25), which was replenished every 48 h.
Preparation of Recombinant SH2 Domain Fusion Proteins-Glutathione S-transferase (GST) fusion proteins were expressed in Escherichia coli and then purified (28). Protein concentration was determined (27) and size and purity evaluated by SDS-PAGE and Coomassie Blue staining. All proteins were Ն95% pure.
Affinity Purification of Interacting Proteins from Cellular Lysates-Interaction of GST-SH2 domains with the GCSF-R was evaluated. We incubated 20 g of GST-Shc SH2 or GST-Grb2 SH2 domain fusion proteins with cellular lysates prepared from BaF3 cells expressing the WT, FFFY, or YYYF GCSF-Rs for 2h at 4°C. Complexes were captured on glutathione-Sepharose 4B by incubation overnight at 4°C, pellets washed, and SDS sample buffer added.
For cells treated with cell-permeable peptides, rapid peptide uptake was confirmed by flow cytometry. 2 Streptavidin-agarose was added to each cellular lysate (3.75 mg of protein/sample) and incubated for 1 h at 4°C. Pellets were washed and SDS sample buffer added.
For evaluation of binding in vitro, each cellular lysate (1.27 mg of protein/sample) was incubated with 3.4 M BP, BP-FENI, or BP-Y*ENI overnight at 4°C. Streptavidin-agarose was added during the final 60 min, after which pellets were washed and resuspended in SDS sample buffer.
Immunoblotting and Far Western Analyses-Proteins (either total cell lysates, or the resulting pelleted proteins from association with the GCSF-R-derived peptides) were separated by SDS-PAGE on 10 or 12% gels as appropriate, transferred to Hybond ECL nitrocellulose membranes, and then blotted with antibodies specific for Shc, Grb2, phosphotyrosine, GCSF-R, or phospho-Stat3. In parallel experiments we also immunoblotted with antibodies to PLC-␥, Stat3, SHP-2, Fyn, or 3BP2, but in these instances either there was no indication of association with the membrane-distal tyrosine (PLC-␥, Stat3, SHP-2, or Fyn) or poor immunoreactivity and poor specificity of reaction with total cell lysates (3BP2). 2 Nitrocellulose membranes were blocked for 1 h in either 1% (w/v) BSA/TBST (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Tween 20) for the detection of phosphotyrosine; 1% (w/v) BSA/TBS for phospho-JNK or phospho-ERK; or 5% (w/v) milk powder/TBST for the detection of all remaining proteins. Membranes were then incubated overnight in the primary antibody (1/500 to 1/2000 in the appropriate block solution or 0.1% (w/v) BSA/TBST for phospho-JNK or phospho-ERK; 4°C). Following incubation with HRP-labeled secondary antibodies, detection was by enhanced chemiluminescence using either the Supersignal reagent or the higher sensitivity LumiLight Plus reagent as appropriate.
In addition, Far Western analysis of binding partners was performed. GST fusion proteins were separated by SDS-PAGE, transferred onto nitrocellulose, and blocked in 5% (v/v) fetal calf serum, 1% (w/v) BSA, 5% (w/v) milk in 50 mM Tris-HCl, pH 7.5, for 1 h. Membranes were washed in Buffer A (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.1 mM dithiothreitol) and incubated overnight at 4°C in 5 M biotinylated Y*ENI (B-Y*ENI) in Buffer A. Membranes were then washed in Buffer A without dithiothreitol and incubated in streptavidin-HRP (1/1000 in 1% (w/v) milk/TBST) for 2 h at room temperature. Following washing in TBST, bound phosphopeptide was detected by enhanced chemiluminescence. The position of the GST-SH2 domain fusion proteins was confirmed by Ponceau Red staining.
BIAcore Analyses-Real-time kinetic studies of the interactions between GST-SH2 domain fusion proteins and peptides based on the membrane-distal tyrosine of the GCSF-R were performed on a BIAcore 2000 biosensor. For all studies, the running and sample dilution buffer was 10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% (v/v) Tween 20. The biotinylated peptides (BP, BP-FENI, and BP-Y*ENI) were immobilized onto 3 flow cells of each 4-flow cell streptavidinbiosensor chip. Low concentrations corresponding to ϳ50 -100 resonance units (RU) were used to control for mass transport effects and the avidity effects of GST dimers (29). The remaining flow cell on each chip remained blank to measure background interactions.
The binding of a GST-SH2 domain protein to the immobilized peptides was recorded as RU in real time to provide a sensorgram. A flow rate of 20 l/min reduced mass transport effects and the overestimation of affinity as a result of rebinding. In these studies, the chip was regenerated with 0.1% (w/v) SDS between injections. Data were analyzed using BIAevaluation software version 3.1.
In-solution competitive binding assays were used to evaluate the relative affinities of the SH2 domains for the phosphopeptides. The GST-SH2 domains at subsaturating amounts (400 nM or 1 M) were preincubated with increasing amounts of the Y*ENI, Y*ENQ, Y*AAA, Y*EAI, or Y*VNV peptides and flowed over the immobilized B-Y*ENI peptide. For these studies, the chip was regenerated with 3 M potassium thiocyanate (two 1-min pulses) at the end of each cycle, followed by 6 M guanidine HCl (two 1-min pulses) at the end of each day. Data (RU expressed as a percentage of total RU bound in the absence of competing peptide) were plotted against the log [competing peptide]. The data were fitted to a Logistic 3 parameter sigmoidal plot by nonlinear regression using Sigma Plot (SPSS Inc.), and IC 50 values determined as the concentration of peptide required to displace 50% of the bound SH2 domain.
JNK Activation Assays-Activation of JNK was assayed by the in vitro phosphorylation of GST-c-Jun (30). Cellular lysate (75 g) was incubated with GST-c-Jun and glutathione-Sepharose overnight at 4°C. Following washing, the kinase reaction was initiated with 1Ci of [␥-32 P]ATP in reaction buffer (20 mM HEPES, pH 7.6, 20 mM MgCl 2 , 20 mM ␤-glycerophosphate) containing 20 M ATP, 100 M sodium vanadate, and 500 M dithiothreitol, then incubated at 30°C for 30 min. 32 P incorporation into the substrate was assessed by autoradiography followed by Cerenkov counting of the SDS-PAGE-separated proteins. For assessment of JNK and ERK activity in the presence of the JNK inhibitor, Tat-TIJIP, lysate (200 g) from G-CSF-stimulated WT GCSF-R BaF3 cells was incubated with 40 g of GST-c-Jun or GST-Elk (26), complexes captured on glutathione-Sepharose 4B and divided into two aliquots. Samples were then treated with buffer or Tat-TIJIP (2 M) for 10 min at 30°C, prior to assaying of kinase activity by incubation with [␥-32 P]ATP as described above.

Evaluation of Cell Survival and Proliferation Mediated by
Mutant GCSF-Rs-To evaluate the contribution of the GCSF-R membrane-distal tyrosine to G-CSF-mediated biological effects, we assessed cell survival and cell proliferation mediated by the wild-type GCSF-R in parallel with GCSF-R mutants containing only the membrane-distal tyrosine (FFFY), without this tyrosine (YYYF), or with all tyrosines mutated (FFFF). The expression of comparable levels of the GCSF-Rs was first confirmed by flow cytometry using LMM741 and an anti-mouse IgG 1 -PE conjugate. 2 For survival studies, cells were deprived of G-CSF for between 2 and 24 h. G-CSF was returned to the media for the remaining time in the 24-h period, and the viability determined by staining with annexin V-FITC and PI at this 24-h time point. Staining was detected by flow cytometry, and those cells excluding both reagents were counted as viable cells (Fig. 1a). No significant changes in viability were observed with up to 5 h of G-CSF deprivation (Fig. 1a, upper left panel). Following 7 and 8 h of G-CSF withdrawal, all of the GCSF-R mutants showed reduced survival. Viability of the YYYF and FFFF GCSF-R mutants continued to fall during the subsequent 8 -12 h of G-CSF withdrawal (Fig. 1a, middle and lower left panels). In addition, after 10 h of G-CSF withdrawal, the survival of cells expressing the wild-type GCSF-R was also compromised ( Fig. 1a, middle right panel). Deprivation of G-CSF for 12-24 h affected the survival of all cells (Fig. 1a, lower panels), and specifically by 24 h there was no significant difference in the survival advantage noted for each cell line following prolonged G-CSF withdrawal (Fig. 1a, lower right panel). These observations confirm an essential role of signaling by the tyrosine residues of the GCSF-R with the loss of all tyrosines of the GCSF-R being most deleterious to cell survival following short periods of G-CSF withdrawal (7-12 h). Loss of the three membrane-proximal tyrosines in the FFFY GCSF-R mutant had less severe effects, whereas loss of the membrane-distal tyrosine alone in the YYYF GCSF-R mutant was intermediate in effect. This suggests that the membrane-distal tyrosine contributes to survival signaling by the GCSF-R.
The contribution of the membrane-distal tyrosine to G-CSFmediated cellular proliferation was also evaluated in longer term experiments in which the different cell lines expressing the WT or YYYF, FFFY, or FFFF mutant GCSF-Rs were cultured for 5-6 days in G-CSF. The viable cell number was quantitated by counting the cells that excluded trypan blue, and the total number of viable cells was used as an estimation of cell proliferation (Fig. 1b). As a control, we first confirmed that the expression of the GCSF-R mutants did not affect proliferation in the presence of IL-3. 2 Furthermore, loss of the membrane-distal tyrosine did not affect proliferation over 5 days in the presence of G-CSF. In contrast, the loss of all tyrosine residues completely prevented cell proliferation. The loss of the three membrane-proximal tyrosines in the FFFY , and FFFF (white) GCSF-Rs were washed and cultured in the absence of G-CSF for the times indicated. G-CSF was then replaced at various times up to 12 h after withdrawal, and viability quantitated 24 h later by co-staining with annexin V and PI, followed by analysis by flow cytometry. The percentage of viable cells was normalized to 100% at the 2-h time point to control for variations in initial viability between cell lines. The results are shown as mean Ϯ range (n ϭ 2 for each clone). b, BaF3 cells expressing the WT (circles), FFFY (squares), YYYF (triangles), and FFFF (diamonds) GCSF-Rs were washed and cultured in the presence of IL-3 (20 ng/ml; Footnote 2) or G-CSF (50 ng/ml) for 5-6 days. Cells were diluted in fresh media every 2 days, and growth factor replenished. Viable cells were quantitated by trypan blue exclusion and related to the initial plating density. Results are shown as mean Ϯ S.E. (n ϭ 3), except FFFF where mean Ϯ range is shown (n ϭ 2). Analysis of variance was performed on the WT, FFFY, and YYYF GCSF-Rs using the Fisher's least squares difference test (p Ͻ 0.05, indicated by *). Similar results were observed in one other clone for WT, FFFY, and YYYF, and two other clones for FFFF. GCSF-R mutant was less severe, with quantitation of its effects revealing that it caused a reduction in cell number to 30% of the wild-type GCSF-R cells after 5 days. Although this latter observation indicates a role for one or more of the membraneproximal tyrosines in mediating cell proliferation, it also suggests that there is some ability of the membrane-distal tyrosine to maintain proliferation above that of the FFFF GCSF-R mutant.
The Membrane-distal Tyrosine of the GCSF-R Is Necessary and Sufficient for JNK Activation-The membrane-distal tyro-sine of the murine GCSF-R has been previously shown through the overexpression of the murine YYYF mutant GCSF-R to be essential for G-CSF-stimulated JNK activation (6). We evaluated JNK activation in BaF3 cell lines stably expressing the wild-type human GCSF-R, or YYYF and FFFY mutant human GCSF-Rs. This showed that JNK activation and phosphorylation by the wild-type human GCSF-R was maximal at 15 min and returned to basal within 60 min. It also confirmed that a human YYYF GCSF-R mutant completely abrogated the JNK activation and phosphorylation (Fig. 2, a and b). This was FIG. 2. Activation of JNK and ERK MAPKs by the GCSF-R. a, BaF3 cells stably transfected with the wild-type and mutant GCSF-Rs were stimulated with G-CSF for 5-60 min, or IL-3 for 15 min. JNK MAPK activation was assayed by in vitro kinase assays using GST-c-Jun as a substrate (upper panels) and phospho-JNK immunoblotting (lower left panels). The positions of the phosphorylated JNK isoforms JNK1 and JNK2 are shown by the closed arrows to the right of these panels. A band that was present in the YYYF experiments, but did not change upon G-CSF or IL-3 stimulation, is shown by the open arrowhead. Equal loading was confirmed by immunoblotting for total JNK1 (lower right panels). These experiments are representative of at least two independent experiments. b, incorporated radioactivity in the JNK MAPK activation assay was quantitated by Cerenkov counting and expressed as the percentage of activation compared with IL-3 stimulation. Results for G-CSF-stimulated cells are indicated: wild-type (circles), YYYF (triangles), FFFY (squares). The results are shown as mean Ϯ S.E. (n ϭ 3 replicates of experiments performed on a single clone) except YYYF (n ϭ 2). c, BaF3 cells stably transfected with the wild-type and mutant GCSF-Rs were stimulated with G-CSF or IL-3 for 15 min. ERK MAPK activation was assayed by in vitro kinase assays using GST-c-Elk as a substrate (upper panels) and phospho-ERK immunoblotting (lower left panels). The positions of the phosphorylated ERK isoforms ERK1 and ERK2 are shown by the closed arrows to the right of these panels. Equal loading was confirmed by immunoblotting for total ERK1 and ERK2 (lower right panels). These experiments are representative of at least three independent experiments. observed for three independent clones of the wild-type and YYYF mutant GCSF-R cells. For the phospho-JNK blots of the YYYF mutant GCSF-R cells shown in Fig. 2a, we also noted the presence of a nonspecific band (see open arrow in Fig. 2a). This was not noted in the other cell lines expressing the wild-type or YYYF mutant GCSF-Rs examined. Thus, the presence of this band was peculiar to this one specific YYYF mutant clone, and its presence did not appear to correlate with the altered JNK activation. Taken together, these results confirm that the loss of the membrane-proximal tyrosine in the human GCSF-R abrogates JNK activation, as might be expected from similar studies of the murine GCSF-R YYYF mutant (6).
We next examined JNK activation following stimulation of a human GCSF-R mutant in which the three most membraneproximal tyrosines were replaced with phenylalanine, thereby creating a FFFY GCSF-R mutant. There was no significant reduction or delay in JNK activation in response to G-CSF when this mutant was stimulated with G-CSF (Fig. 2, a and b). The activation of JNK following IL-3 stimulation was not affected by GCSF-R mutation (Fig. 2a). None of the GCSF-R mutations affected the rapid activation of the kinase activity of ERK assessed in the Elk kinase assay or the phosphorylation of ERKs as observed by immunoblotting. The activation of ERK following IL-3 stimulation was likewise unaffected (Fig. 2c). Therefore, the membrane-distal tyrosine of the GCSF-R appears to be both necessary and sufficient for JNK activation by G-CSF, but is not involved in the control of ERK activation.
JNK Activation Contributes to Proliferation Signaling by the Membrane-distal Tyrosine of the GCSF-R-The membrane-distal tyrosine contributed to G-CSF-stimulated BaF3 cell proliferation (Fig. 1b) and promoted JNK activation (Fig. 2). To further evaluate the biological role of JNK activation following G-CSF stimulation, we utilized a peptide inhibitor of JNK to determine the contribution of JNK to proliferation by the membrane-distal tyrosine. The peptide TIJIP has been derived from the scaffold protein JIP-1, and selectively binds and inhibits the in vitro activity of activated JNK, but not p38 or ERK MAPK (25). To confirm that a cell-permeable version of this peptide, Tat-TIJIP, retained specific JNK inhibition, the in vitro activity of JNK and ERK following G-CSF-stimulation of BaF3 cells was assessed in the presence of Tat-TIJIP (Fig. 3a). The stimulation of BaF3 cells stably expressing human wildtype GCSF-R with G-CSF resulted in maximal JNK activity in vitro (Fig. 3a, left panel, lane and column 3). In the presence of 2 M Tat-TIJIP, JNK activity was reduced by more than 50% (Fig. 3a, left panel, lane and column 4; p Ͻ 0.05), whereas Tat-TIJIP did not affect basal JNK activity (Fig. 3a, left panel,  lanes and columns 1 and 2). In contrast, Tat-TIJIP did not inhibit the activity of ERK toward the substrate GST-Elk following G-CSF-stimulation (Fig. 3a, right panel), indicating Tat-TIJIP specifically inhibits JNK, but not ERK activity in vitro.
To further investigate how the membrane-distal tyrosine might contribute to proliferation of the FFFY mutant above that of the FFFF mutant (Fig. 1b), we utilized the Tat-TIJIP peptide inhibitor of JNK. BaF3 cells overexpressing the FFFY mutant GCSF-R were cultured in G-CSF for 6 days in the presence of Tat-TIJIP and viable cell number quantitated as an estimation of proliferation as before (Fig. 3b). The addition of 10 M Tat-TIJIP further attenuated the remaining proliferation of FFFY GCSF-R BaF3 cells in G-CSF, indicating a role for the JNK pathway initiated from the membrane-distal tyrosine of the GCSF-R in cellular proliferation by G-CSF.
The Full-length Wild-type and FFFY GCSF-Rs Interact with the Shc SH2 Domain in Vitro-The membrane-proximal pathways leading to JNK activation remain controversial, with various adaptors (31,32), small G-proteins (33), and protein kinases (34) being implicated. The requirement for the membrane-distal tyrosine in the GCSF-R mediated JNK activation as previously reported (6) suggests that this system provides a useful model to evaluate the signaling events initiating JNK activation following cytokine stimulation. As a first step in this evaluation, we used the data provided by Songyang and colleagues (20) to predict binding partners for the peptide sequences surrounding the membrane-distal tyrosine of the GCSF-R. Multiple partners including the adaptor proteins Grb2 and Shc were predicted. 2 Indeed, earlier experimental data by Ward and colleagues (13,35) suggested the potential for the GCSF-R membrane-distal tyrosine to act as a docking site for multiple SH2 domain-containing proteins such as the adaptors Grb2 and the tyrosine kinase Hck. Furthermore, the tyrosine phosphorylation of Shc following G-CSF stimulation has been reported to be mediated predominantly by this distal tyrosine (6,14,36). Therefore, in the following experiments, we first focused on experimentally determining direct binding partners for the phosphorylated membrane-distal tyrosine of the GCSF-R.
Our initial attempts to co-immunoprecipitate the full-length GCSF-R and SH2 domain-containing proteins, and thereby determine their interaction in intact hematopoietic cells, were hampered by the low levels of GCSF-R expression and the variable efficacy of commercially available antibodies to detect and/or immunoprecipitate SH2 domain-containing proteins or the GCSF-R. We therefore adopted a strategy in which cell lysates were incubated with GST fusion SH2 domains. The isolation of the fusion proteins, and subsequent immunoblotting thus allowed the detection of SH2 domain-containing proteins with the potential to bind the wild-type GCSF-R. As shown in the upper two panels of Fig. 4a, binding of the total GCSF-R by both Shc and Grb2 SH2 domains was detected. We found that longer exposure times were required to detect the binding to the Grb2 SH2 domain. We confirmed binding was mediated by the phosphorylated GCSF-R (Fig. 4a, lower  panels).
We then tested the ability of GST-SH2 domains to interact with a GCSF-R mutant in which only the membrane-distal tyrosine was intact. This mutant, the so-called FFFY GCSF-R mutant, allowed us to evaluate binding without the complication of contributions by other tyrosines in the full-length wildtype receptor (Fig. 4b). By incubating cellular lysate with the GST-Shc SH2 (lanes P1 and P2) or GST-Grb2 SH2 (lanes P3 and P4), we could affinity-purify the FFFY GCSF-R mutant only with the GST-Shc SH2 domain following stimulation with G-CSF (lane P2). Longer exposure times failed to detect interaction with the Grb2 SH2 domain. 2 This suggests a robust interaction of the Shc SH2 domain with the membrane-distal tyrosine of the GCSF-R. It also demonstrates the SH2 domain of Shc interacts with this phosphotyrosine in the context of the full-length receptor.
Finally, the interaction of the Shc and Grb2 SH2 domains with the YYYF GCSF-R mutant was tested. The YYYF mutant could not be purified by either the Shc (lanes P1 and P2) or Grb2 SH2 (lanes P3 and P4) domains, even after G-CSF stimulation (Fig. 4c). This suggests that the Shc and Grb2 SH2 domains do not interact with the three membrane-proximal tyrosines of the GCSF-R, and Shc binds solely to the membrane-distal tyrosine. In agreement with previous reports (14,36,37), we also found that the tyrosine phosphorylation of Shc following G-CSF stimulation was abolished in the YYYF GCSF-R mutant. 2 This supports a model in which Shc interacts directly or indirectly with the membrane-distal tyrosine in the context of the full-length GCSF-R and itself becomes tyrosine- The Phosphorylated GCSF-R Peptide, BP-Y*ENI, Binds to Shc When Delivered Intracellularly-Because our comparison between wild-type and mutant GCSF-R-expressing cell lines may be complicated by small differences in the levels of GCSF-R expression, we investigated approaches to study interactions in vitro that could use peptides based on the region surrounding the GCSF-R membrane-distal tyrosine. We focused on the interactions of Shc or Grb2 SH2 domain interactions with these peptides. First, to evaluate the interactions in the context of the intact cell, we delivered a cell-permeable version of a peptide sequence surrounding the membrane-distal tyrosine of the GCSF-R intracellularly. Specifically, the penetratin sequence has been derived from the third helix of the Antennapedia homeoprotein and it crosses cell membranes in a receptor-independent manner (38,39). We utilized this property to deliver the BP-Y*ENI peptide. By flow cytometry with detection using streptavidin-FITC, we estimated that 94% of cells took up this peptide within 5 min of treatment. 2 Following treatment of cells with BP-Y*ENI for 3 h prior to lysis, it was possible to affinity-purify Shc from the cell (Fig. 5a,  upper panel, lane P4). Shc could not be co-purified with the non-phosphorylated phenylalanine peptide (BP-FENI) or the penetratin vector alone (BP; Fig. 5a, upper panel, lanes P2 and  P3) thereby emphasizing the specificity of this interaction. Also under these assay conditions, it was not possible to detect an interaction of Grb2, Fyn, PLC-␥, SHP-2, or Stat3 with the BP-Y*ENI peptide (Fig. 5a, middle and lower panels). 2 These results suggest that, in the intracellular environment, Shc appears to be a preferred binding partner for the membranedistal tyrosine of the GCSF-R.
Binding Assays Detect Binding of Shc and Also Grb2 to the Phosphorylated GCSF-R Peptide, BP-Y*ENI-We next incubated BP-Y*ENI directly with cell lysates to minimize any potential problems of intracellular accessibility. Following overnight incubation of lysates with BP-Y*ENI at 4°C, fulllength Shc as well as Grb2 were isolated from the cellular lysates (Fig. 5b, upper and middle panels, lanes P4). The specificity of this interaction was confirmed by demonstrating that penetratin alone (BP), as well as an unphosphorylated peptide BP-FENI, was unable to interact with Shc or Grb2 (Fig. 5b,  lanes P2 and P3 in each panel). We also demonstrated that there was no association with another SH2 domain-containing protein PLC-␥ and very limited association with the Src family member Fyn (Fig. 5b, lower panel). 2 Thus, in this in vitro assay, the BP-Y*ENI peptide specifically associated with Shc and Grb2, allowing these adaptor proteins to be partially purified from a complex mixture of proteins in a cellular lysate.
We also utilized a simpler in vitro Far Western analysis in which purified GST-SH2 domains were separated by SDS-PAGE. The proteins were transferred, and the resulting membrane was probed with a biotinylated Y*ENI peptide (B-Y*ENI) and then streptavidin-HRP. An interaction was also detected with the Shc and Grb2 SH2 domains (Fig. 5c, upper and middle panels). No interaction was observed for GST-SH2 domain of Fyn, Src, or PLC-␥ (Fig. 5c, lower panel). 2 Identification of SH2 Domain-containing Proteins as Binding Partners for the Phosphorylated GCSF-R Peptide, BP-Y*ENI, by BIAcore Biosensor-In the preceding experiments, we consistently observed the binding of the Shc SH2 domain to the membrane-distal tyrosine of the GCSF-R, or peptides based on this motif. The binding by the Grb2 SH2 domain was only observed by lysate affinity purification and Far Western analysis (Fig. 5, b and c). However, these methods all employ stringent wash conditions, which may compromise the observation of lower affinity or lower abundance binding partners. We therefore moved to define more rigorously the binding interactions of the SH2 domain-containing proteins with the BP-Y*ENI peptide using BIAcore analysis. In this way we could measure, in real time, the interaction of native SH2 domain-containing proteins with the BP-Y*ENI peptide, without artifacts potentially introduced through separation of bound and free proteins in the washing protocols, the denaturloading of substrates was confirmed by Coomassie staining (see Footnote 2). JNK and ERK activation is shown expressed as a percentage of maximal activation by G-CSF-stimulation (lower panels). Analysis of variance was performed using the Fisher's least squares difference test, where the effect of addition of Tat-TIJIP was evaluated in unstimulated (Ϫ) and G-CSF-stimulated (ϩ) cells (p Ͻ 0.05, indicated by *). b, BaF3 cells expressing the FFFY GCSF-R were washed and cultured for 6 days in either G-CSF (50 ng/ml) or G-CSF and 10 M Tat-TIJIP. Cells were diluted in fresh media every 2 days, and G-CSF and Tat-TIJIP replenished. Viable cells were quantitated by trypan blue exclusion and related to the initial plating density. The percentage of viable cells following treatment with Tat-TIJIP, compared with that in the absence of inhibitor, is shown as mean Ϯ S.E. (n ϭ 3). Analysis of variance was performed using the Fisher's least squares difference test (p Ͻ 0.05, indicated by *; p Ͻ 0.01, indicated by **). This was repeated using another independent FFFY GCSF-R BaF3 clone with similar results (see Footnote 2). ation of proteins in Far Western protocols, or prolonged incubations at lowered temperatures.
We screened SH2 domains from 12 different SH2 domaincontaining adaptor proteins and enzymes for selective binding to the phosphotyrosine peptide BP-Y*ENI. In this approach, we continued to use the BP-Y*ENI peptide because the peptide sequence of penetratin would hold the peptides away from the biosensor chip and ensure access to the immobilized peptide sequence. Using this technique we found that, within the adaptor class of proteins, the SH2 domains of Shc and Grb2, as well as the novel adaptor 3BP2, exhibited phosphotyrosine-dependent binding to BP-Y*ENI (Fig. 6a). There was very little nonspecific interaction of the GST-SH2 domains with the nonphosphorylated control peptide BP-FENI or the blank surface of the chip. 2 In contrast, the SH2 domains of the adaptors Nck and the p85 subunit of phosphatidylinositol 3-kinase did not interact. 2 Of the enzymes screened, the SH2 domains of the Src family kinases Hck and Fyn bound BP-Y*ENI (Fig. 6a). Under these assay conditions, it was not possible to detect any interaction of SH2 domains from Src, Syk, ZAP70, SHP-2, RasGAP, or GST with the BP-Y*ENI peptide (Fig. 6a). 2 Syk and SHP-2 fusion proteins each contain tandem SH2 domains (40,41), but when we tested constructs containing both N-and C-terminal SH2 domains, we also failed to show their interaction with the BP-Y*ENI peptide (Fig. 6a). 2 These results provide further evidence that the membrane-distal tyrosine can potentially interact with two adaptors (Shc and Grb2) implicated originally in the ERK pathway (34,42), as well as a novel adaptor 3BP2, which mediates interactions with tyrosine kinases such as Syk and Abl (43,44).
Kinetic Analysis of Binding Partners of Peptides based on the Membrane-distal Tyrosine-Given that we identified five proteins as potential binding partners for the phosphorylated GCSF-R membrane-distal tyrosine residue, we then utilized the analytical capabilities of the BIAcore analysis system to quantitate the binding of these proteins. We therefore deter-  Footnote 2). The sizes of these proteins, as calculated by migration relative to molecular mass markers, are shown to the right of each panel. c, the phosphorylated peptide based on the membrane-distal tyrosine of the GCSF-R was used in Far Western analyses. The SH2 domains of Shc (upper panel), Grb2 (middle panel), Fyn (lower panel) were produced as GST fusion proteins. Each protein was serially diluted (10 g to 0.1 g) and transferred onto nitrocellulose. GST only (10 g) was included as a negative control, and we also could not detect interaction with the SH2 domains from PLC-␥ or Src (see Footnote 2). Probing with 5 M B-Y*ENI was detected with streptavidin-HRP and enhanced chemiluminescence using the Supersignal reagents. The arrows to the right of each panel indicate the position of the GST fusion proteins estimated from Ponceau Red staining of the nitrocellulose membranes. mined the relative IC 50 values for interaction of the SH2 domains for the Y*ENI peptide by an in-solution competition assay (29,45). The IC 50 values for the SH2 domains of the adaptors Shc, Grb2, and 3BP2 were in the low micromolar range (Fig. 6b, right-hand panel). A full-length Grb2 protein displayed a similar IC 50 value for interaction. 2 Using this approach, we estimated the IC 50 values for interaction of Fyn and Hck SH2 domains ranged between 10 and 15 M (Fig. 6b,  right-hand panel). These results essentially confirm the predictions of SH2 binding preferences made from the binding data of Songyang and colleagues (18,20). Thus, Shc (with a preference for Y*(E/I)X(I/L/M)), Grb2 (with a preference for Y*((Q/Y)NY), 3BP2 (with a preference for Y*ENX), and Fyn and Hck (with a preference for Y*EEI) all bind the Y*ENI motif based on the membrane-distal tyrosine of the murine GCSF-R.
The SH2 Domain of the Novel Adaptor 3BP2 Interacts with the Full-length GCSF-R in a Phosphotyrosine-dependent Manner-The SH2 domain of the relatively poorly characterized adaptor, 3BP2, interacted with the Y*ENI peptide. The IC 50 value was comparable with that of the Shc and Grb2 SH2 domains, which have been previously implicated in interaction with the membrane-distal tyrosine of the GCSF-R and/or signaling following stimulation with G-CSF (6,13,14,36). We have been able to demonstrate a phosphotyrosine-dependent interaction between the Shc and Grb2 SH2 domains and the full-length GCSF-R (Fig. 4a). To determine whether 3BP2 may be a biologically relevant binding partner for the GCSF-R, we assessed the ability of the 3BP2 SH2 domain to affinity-purify the full-length wild-type GCSF-R from a cellular lysate (Fig. 6,  c and d). This indicated that the 3BP2 SH2 domain was able to interact with the full-length wild-type GCSF-R (Fig. 6c), and that this interaction was dependent on tyrosine phosphorylation of the GCSF-R (Fig. 6d). As observed with the Grb2 SH2 domain (Fig. 4a), the FFFY and YYYF GCSF-Rs could not be detected by immunoblotting with the anti-GCSF-R antibody LMM741 following incubation of lysate with GST-3BP2 SH2. 2 This confounding observation may again reflect the diminished sensitivity in this assay because of slight variations in GCSF-R expression levels, especially the observed lower levels of GCSF-R FFFY expression. 2 Altering the Sequence Context of Peptides Allows Evaluation of the Requirement for Direct Interaction of Shc with the Membranedistal Tyrosine for JNK Activation-The binding preferences of SH2 domains have been experimentally shown to depend on the amino acid sequence immediately C-terminal of the phosphorylated tyrosine (17,46). This provides an additional control showing the specificity of binding to the BP-Y*ENI peptide sequence.  It also provides a way to evaluate the contributions of the potential binding partners Shc, Grb2, and 3BP2 to signaling initiated by the GCSF-R membrane-distal tyrosine.
The observations in this and previous studies suggest G-CSF stimulates an interaction of Shc with the membrane-distal tyrosine of both the murine and human GCSF-Rs (Figs. 4 and 5; Refs. 6, 14, 36, and 37). To test whether the direct recruitment of Shc to the phosphorylated membrane-distal tyrosine was required for JNK activation, the contribution of the sequence context of the phosphotyrosine to SH2 domain binding was exploited. Specifically, the three amino acids immediately C-terminal to the tyrosine residue were altered so that the new sequence was predicted to have a low affinity for interaction with Shc. The sequence was thus changed from YENI to YENQ. We then determined the IC 50 values for in vitro binding of the Shc, 3BP2, and Grb2 SH2 domains to the CSPKSY*ENQWamide (termed Y*ENQ) peptide. This revealed the IC 50 values for 3BP2 and Grb2 SH2 interaction were relatively unchanged compared with the Y*ENI peptide (Fig. 7a). In contrast, interaction of the SH2 domain of Shc was greatly reduced as predicted (Fig. 7a).
Next, the YENQ sequence context mutation was created within the full-length human GCSF-R, and this was expressed in BaF3 cells. The human GCSF-R was chosen so that the human GCSF-R-specific antibody LMM741 could be used for cell sorting. When JNK activation by human wild-type and YENQ mutant GCSF-Rs was examined, the YENQ GCSF-R, which would be expected not to bind Shc (Fig. 7a), did not prevent JNK activation (Fig. 7b). This suggested Shc was not essential for JNK activation mediated by the membrane-distal tyrosine of the GCSF-R.
The SH2 domain of Stat3 shows high affinity binding to YXXQ motifs, including that of the membrane-proximal tyrosine of the GCSF-R (19). Thus, the alteration of the sequence context of the membrane-distal tyrosine to YENQ could potentially create an additional Stat3 binding site within the intracellular region of the GCSF-R. To address the effect of the YENQ mutation on G-CSF-mediated Stat3 Tyr-705 phosphorylation, lysates were prepared from human wild-type, YENQ, YYYF, and FFFY GCSF-R BaF3 cells, and immunoblotted for phosphoStat3 (Fig. 7c). Following G-CSF stimulation of the wild-type GCSF-R, Stat3 tyrosine phosphorylation was rapid, FIG. 7. JNK activation by a fulllength GCSF-R membrane-distal tyrosine mutant that prevents direct Shc binding. a, in-solution competition assays were utilized to assess the affinities of the SH2 domains for the GCSF-R membrane-distal tyrosine within the altered sequence context of YENQ (rather than the YENI of the wild-type GCSF-R sequence context). A constant, subsaturating amount (400 nM) of each GST fusion protein was preincubated for 5 min on ice with increasing amounts of the competing peptide Y*ENQ. IC 50 values were then calculated for Shc SH2 (closed circles), 3BP2 SH2 (closed triangles), and Grb2 SH2 (open circles), and these values are shown to the right of each panel. To control for differences introduced by measuring IC 50 values for different peptides on different days, the binding of GST-Shc SH2 to Y*ENQ was evaluated on 1 day, and then the binding of Shc was directly compared with each of the peptides on the same day and used to normalize the values. b, BaF3 cells stably transfected with the wild-type and YENQ mutant human GCSF-Rs were stimulated with G-CSF for 5-60 min, or IL-3 for 15 min. JNK MAPK activation was assayed by in vitro kinase assay using GST-c-Jun as a substrate, quantitated by Cerenkov counting and expressed as the percentage of activation compared with IL-3 stimulation. Results for G-CSF-stimulated cells are indicated: wild-type (closed circles) and YENQ (upside-down triangles). The results are shown as mean Ϯ S.E. (n ϭ 3). c, Stat3 tyrosine phosphorylation (Tyr-705) was assessed by stimulation of wildtype (upper panel), YENQ (second panel), YYYF (third panel), and FFFY (lowest panel) BaF3 cells with G-CSF for the times indicated, and preparation of lysates. Following SDS-PAGE, membranes were probed with phospho-Stat3 antisera. Equal loading was confirmed by stripping membranes and reprobing for total Stat3 (see Footnote 2). occurring within 5 min, and was sustained for at least 60 min (Fig. 7c, upper panel). Stimulation of the YENQ GCSF-R mutant with G-CSF resulted in a similarly rapid and sustained Stat3 tyrosine phosphorylation (Fig. 7c, second uppermost  panel), suggesting the YENQ mutation did not alter Stat3 tyrosine phosphorylation by the GCSF-R. In addition, the requirement for the membrane-distal tyrosine or the three membrane-proximal tyrosines of the GCSF-R in Stat3 tyrosine phosphorylation was assessed utilizing the YYYF and FFFY mutant GCSF-Rs (Fig. 7c, lower panels). Both the YYYF and FFFY GCSF-Rs retained Stat3 tyrosine phosphorylation, suggesting that neither the membrane-distal tyrosine nor the three membrane-proximal tyrosines are essential for mediating Stat3 tyrosine phosphorylation. This is consistent with the report that Stat3 activation in BaF3 cells was independent of GCSF-R intracellular tyrosines at saturating concentrations of G-CSF (47), as used in this study.
The Contribution of the Sequence Context of Peptides Based on the GCSF-R Membrane-distal Tyrosine to Binding Affinities-Alteration of the sequence context of the membrane-distal tyrosine of the human GCSF-R to YENQ did not prevent JNK activation (Fig. 7b). This suggested interaction of the membrane-distal tyrosine with Shc was not required for JNK activation by this site, and raised the possibility that the other identified partners 3BP2 and Grb2 may be required. Thus, the effect of three additional GCSF-R sequence context alterations on the in vitro binding of Shc, 3BP2, and Grb2 and JNK activation were assessed. A role for Fyn or Hck in G-CSFmediated JNK activation was discounted using the high affinity Src family tyrosine kinase inhibitor PP2 2 (IC 50 ϭ 5 nM) (48); therefore, they were not examined further. As shown in Fig. 8a, changing the sequence from CSPKSY*ENIW-amide to CSPKSY*AAAW-amide (termed Y*AAA from here on) did not change binding to Shc but attenuated binding to Grb2 and 3BP2. Changing the sequence to CSPKSY*EAIW-amide (termed Y*EAI) did not change Shc or 3BP2 binding but diminished binding to Grb2 (Fig. 8b); changing the sequence to CSPKSY*VNVW-amide (termed Y*VNV) did not abrogate binding to Shc, Grb2, or 3BP2 (Fig. 8c). These results are largely consistent with the predictions made by Songyang and colleagues (20). The exceptions are that the Y*VNV and Y*AAA peptides retain Shc binding despite its predicted preference for binding the sequence Y*(E/I)X(I/L/M).
Modification of the Sequence Context of the Membrane-distal Tyrosine of the GCSF-R Abrogates JNK Activation-To evaluate how the binding partners for the GCSF-R membrane-distal tyrosine alter JNK activation following exposure to G-CSF, we created cell lines expressing modified GCSF-Rs. As the Y*AAA peptide retained binding of Shc only, we were able to perform the converse experiment to that utilizing the YENQ peptide, to again investigate the requirement for direct binding of Shc in JNK activation. Because the membrane-distal tyrosine solely mediates JNK activation by both the murine and human GCSF-Rs ( Fig. 2; Ref. 6) and because the murine and human GCSF-Rs have 60.2% identity and 69.6% similarity in their intracellular signal transduction domains, we took the opportunity to confirm that mutation of the murine GCSF-R membrane-distal tyrosine affected JNK activation by creating the sequence context alterations within the murine GCSF-R and expressing these in BaF3 cells.
Following G-CSF stimulation, the YEAI and YVNV mutants retained full JNK activation. However, the YAAA mutant abrogated JNK activation following G-CSF exposure (Fig. 8d). Thus, a sequence that retains the ability to bind Shc but not Grb2 or 3BP2 (Fig. 8a) is unable to fully activate JNK. This confirms that direct binding of the Shc SH2 domain to the GCSF-R is not critical for JNK activation. Furthermore, the lack of binding of Grb2 to the YEAI mutant (Fig. 8b), but the retaining of JNK activation by the YEAI mutant GCSF-R, also suggests that the direct binding of Grb2 to the GCSF-R is not essential. It may therefore be that the less well characterized SH2 domain-containing adaptor 3BP2 mediates JNK activation. Alternatively, other SH2 domain-containing proteins may perform this role, and this is an area requiring characterization in the future. To date, our library screening protocols with B-Y*ENI peptides have not isolated specific binding partners. 2

DISCUSSION
The biological role of cytokine-stimulated JNK activation in hemopoietic cells was investigated in BaF3 cells expressing wild-type or mutant GCSF-Rs. JNK activation has been reported to be involved in a diverse array of seemingly conflicting responses in hemopoietic and other cell types, including survival and apoptosis (49). Loss of the membrane-distal tyrosine of the GCSF-R resulted in reduced survival of BaF3 cells following G-CSF withdrawal (Fig. 1a). This raises the possibility that JNK activation mediated via the membrane-distal tyrosine acts to promote the survival of BaF3 cells following G-CSF withdrawal. This represents the first report of a role for the membrane-distal tyrosine of the GCSF-R in survival.
The role of JNK activation in BaF3 cells was further investigated in the continued presence of G-CSF. In this case, loss of the three membrane-proximal tyrosines resulted in diminished long term proliferation in G-CSF, whereas the FFFF GCSF-R mutant did not support proliferation (Fig. 1b). This is consistent with several previously reported observations utilizing single Tyr 3 Phe GCSF-R mutations in myeloid cell types. Ward et al. (13) reported the involvement of two membrane-proximal tyrosines, as well as Tyr-764, in proliferation of 32D myeloid cells, whereas de Koning and colleagues (36) found the membrane-distal tyrosine markedly influenced both 32D proliferation and timing of differentiation. Furthermore, the membrane-distal tyrosine was implicated in proliferation of primary myeloid cells (12). However, studies utilizing BaF3 cells expressing truncated GCSF-Rs have reported the membranedistal tyrosine to lie within a growth inhibitory domain and to down-regulate proliferation by recruitment of SHIP via Shc (10). Although we have observed complex formation of Shc: SHIP mediated by the membrane-distal tyrosine, 2 our comparison of the FFFY and FFFF GCSF-R mutants revealed that the membrane-distal tyrosine acts in a pro-proliferative rather than an inhibitory manner (Fig. 1b).
The activation of the JNK pathway by G-CSF has been previously reported to require the GCSF-R membrane-distal tyrosine (6). We further show that the requirement of the membrane-distal tyrosine for JNK activation is retained by the human GCSF-R, and that JNK activation is mediated solely via the membrane-distal tyrosine (Fig. 2, a and b). Inhibition of JNK activity with Tat-TIJIP almost completely abrogated the proliferation stimulated by the FFFY GCSF-R, revealing a role for JNK activation initiated from the membrane-distal tyrosine of the GCSF-R in proliferation of BaF3 cells (Fig. 3b). This is consistent with the role for JNK in IL-3-mediated BaF3 proliferation (50).
G-CSF stimulation rapidly leads to the tyrosine phosphorylation of the GCSF-R and other intracellular signaling proteins (51). The phosphotyrosine residues of the GCSF-R then act as high affinity binding sites for specific SH2 domain-containing proteins (16,52). In an attempt to understand the mechanisms by which the membrane-distal tyrosine promotes proliferation and survival in response to G-CSF, we identified potential binding partners for this residue. We observed phosphotyrosinedependent binding of the Shc and Grb2 SH2 domains to the full-length wild-type GCSF-R and the interaction of the Shc SH2 domain with the FFFY mutant GCSF-R (Fig. 4, a and b). With a series of peptides based on the region surrounding the GCSF-R membrane-proximal tyrosine, we also showed the SH2 domains of the adaptors Shc and Grb2 to be direct binding partners (Fig. 5, b and c). These results largely confirm the results of Ward and colleagues (13,35), but differ in that we could not detect binding of the N-terminal SH2 domain of SHP2 to these peptides. 2 We further extended our analysis with the delivery of peptides intracellularly using the penetratin cellpermeable peptide sequence (38,53). With this strategy, which allowed evaluation of interactions within the intracellular environment, we confirmed Shc as a binding partner for this specific tyrosine residue (Fig. 5a).
The comparison of binding to different SH2 domain-containing proteins in vitro, as shown in Fig. 5, can be compromised by differences in antibody affinity/signal strength, the stringent washing protocols, or the intracellular accessibility of the peptides to the SH2 domain-containing proteins. Some of these complications may underlie the differences in binding partners identified by different approaches (13,14,35,37). To avoid these complications, we carried out further in vitro binding assays utilizing the BIAcore biosensor. This allowed rapid screening of 12 different proteins as well as the evaluation of their relative affinities of binding.
In addition to confirming the binding of Shc and Grb2, we also identified 3BP2, a less well characterized adaptor protein (43,54), as a binding partner (Fig. 6a). Of the enzymes tested, Src family tyrosine kinases Fyn and Hck showed some, but limited interaction, but we still could not detect interaction with the SH2 domains of SHP2, despite testing four independent fusion proteins. 2 To determine the relative affinities of binding, we next utilized an in-solution competition assay to provide an estimation of binding affinities (45). The SH2 domains of the adaptor proteins Shc, Grb2, and 3BP2 displayed a similar high affinity for Y*ENI, being in the low micromolar range (Fig. 6b). These agree with affinities previously reported for other optimal SH2 domain-phosphotyrosine interactions (55). Other SH2 domains tested showed affinities of at least an order of magnitude lower for the BP-Y*ENI peptide. 2 Based on these estimated affinities (Fig. 6b), we proposed that the adaptor proteins Shc, 3BP2, and Grb2 can interact with the phosphorylated membrane-distal tyrosine of the GCSF-R.
Although this identification of the membrane-distal tyrosine as a putative multifunctional binding site is of interest, it also brings with it complications in delineating the upstream events leading to JNK activation mediated by this site. Specifically, the mutation of this tyrosine to phenylalanine within the context of the GCSF-R would equally abolish binding to Shc, Grb2, 3BP2, or other, as yet unidentified SH2 domain-containing partners. This loss of binding is observed when these SH2 domains bind Y*ENI peptides but not the equivalent FENI peptides (Fig. 5). Likewise, the overexpression of the SH2 domains of Shc, Grb2, or 3BP2 to act as dominant-negative mutants would all be predicted to sequester the GCSF-R membrane-distal tyrosine following its phosphorylation. This would not allow discrimination of which SH2 domain-containing proteins mediate JNK pathway activation.
Given that the selectivity of SH2 domains for specific sequence contexts has been previously predicted by degenerate phosphopeptide library screening (20), we proposed that altering the residues surrounding the membrane-distal tyrosine of the GCSF-R would provide a useful alternative strategy to delineate which SH2 domain-containing proteins mediate JNK activation. Specifically, we showed that the alteration of the sequence of the Y*ENI peptide to Y*ENQ resulted in the specific loss of Shc binding (Fig. 7a). Next, we used our data on altered binding obtained by BIAcore evaluation in conjunction with JNK activation data when the YENQ change was introduced into the full-length GCSF-R expressed in BaF3 cells. When YENQ mutant GCSF-R BaF3 cells were stimulated with G-CSF, JNK activation was observed, suggesting the direct binding of Shc to the membrane-distal tyrosine was not required for JNK activation (Fig. 7b). Thus additional sequence context mutations were required to characterize the roles of 3BP2 and Grb2 in JNK activation.
As summarized in Fig. 8, the alteration of the Y*ENI peptide to Y*EAI resulted in the loss of Grb2 binding. The Y*AAA peptide lost binding of both 3BP2 and Grb2, whereas interaction of all three adaptors was retained with the Y*VNV peptide. Our observations agree with many of the predictions of the previous phosphopeptide library screening (18,20), with the exceptions that Y*VNV and Y*AAA retain binding of Shc. Such differences emphasize that the library screening has not directly tested specific peptide sequences, but rather has predicted the preferred amino acid at each site (18,20). Furthermore, we demonstrated that amino acid substitutions would alter the affinity of the SH2 domains for the peptide, thereby reconfirming that SH2 binding specificity, even to short peptides, is dependent on the amino acids immediately C-terminal to the phosphotyrosine.
When the sequence immediately C-terminal to the tyrosine was altered in the full-length GCSF-R to modify the repertoire of interacting proteins and thus JNK activation, we found that only the YAAA GCSF-R mutant compromised JNK activation (Fig. 8a). Because the YEAI GCSF-R mutant retains JNK activation, but the Y*EAI peptide can no longer bind Grb2, it would seem that direct Grb2 recruitment to the GCSF-R membrane-distal tyrosine is dispensable for JNK activation. Similarly, because the Y*AAA peptide retained binding to Shc, this would suggest that direct recruitment of Shc to the GCSF-R membrane-distal tyrosine is not essential.
This observation that Shc does not mediate JNK activation by the GCSF-R is not consistent with the findings of Hashimoto et al. (31), where the relative contributions of Shc and Grb2 to JNK activation following stimulation of the epidermal growth factor (EGF) receptor were addressed. EGF-stimulated JNK activation was assessed in cells deficient in Shc or Grb2 expression (Shc or Grb2 knock-out chicken DT40 B cell lines), and was found to require the presence of Shc (31). However, the activation of JNK was far less robust in DT40 cells (2.5-fold) than that observed in BaF3 cells in this study (Ͼ20-fold) and has not been shown to be mediated solely by one tyrosine of the EGF receptor, and the role of direct recruitment of Shc to the EGF-R was not assessed. Thus, different receptors may use different adaptors to initiate JNK activation.
It remains possible that JNK activation is mediated via interaction of a novel or other unidentified signaling protein.
For example, we also identified a new potential candidate for binding, the adaptor protein 3BP2. The SH2 domain of 3BP2 did not bind the Y*AAA peptide (Fig. 8a), and the YAAA mutant GCSF-R was unable to activate JNK (Fig. 8d). As noted previously, the role of 3BP2 in signaling is relatively poorly characterized. 3BP2 is selectively expressed in hematopoietic cells (43,44), and this is consistent with a possible role in mediating signaling events in hematopoietic cells. The 3BP2 SH2 domain has been demonstrated to interact with a number of signaling proteins including intracellular proteins (such as Syk family kinases, LAT, ZAP 70, Grb2, PLC-␥1, and Cbl) and membrane proteins (such as the CR2 receptor and p95) (43,54). Thus, 3BP2 could aid in the assembly of a protein complex at the plasma membrane that leads to JNK activation. The role of 3BP2 in the response to G-CSF therefore requires further investigation; however, this is complicated by the lack of a commercially available antibody of sufficient specificity. 2 Furthermore, a GCSF-R membrane-distal tyrosine mutation to specifically prevent 3BP2 binding could not be tested, as comparison of the 3BP2 SH2 domain binding motif predictions suggest that changes to prevent 3BP2 binding would also prevent Grb2 binding. A remaining strategy would therefore be to test GCSF-R-mediated JNK activation in cells of a 3BP2 knockout animal when these are available. Thus in conclusion, these observations suggest that the membrane-distal tyrosine of the GCSF-R directly recruits signaling intermediates that act to prolong cell survival, and induce pro-proliferative signaling via activation of the JNK MAPK pathway through a mechanism that does not appear to directly involve Shc or Grb2.