A systematic scan of interactions with tyrosine motifs in the erythropoietin receptor using a mammalian 2-hybrid approach

The glycoprotein hormone erythropoietin (Epo) is essential for red blood cell development because Epo^-/- and Epo receptor-deficient (EpoR^-/-) mice die at an embryonic stage due to severe anemia.¹  Binding of Epo to its receptor prevents apoptosis of erythroid progenitors and promotes proliferation and erythroid maturation. The EpoR is a member of the type I cytokine receptor family²  characterized by a single ligand-binding cytokine receptor homology domain³  containing conserved cysteines and a Trp-Ser-X-Trp-Ser motif,⁴  and lacking any catalytic activity in the intracellular region.²  Epo binding to its receptor induces a conformational change of the preformed receptor-dimer resulting in activation and cross-phosphorylation of associated Janus kinase 2 (JAK2) tyrosine kinases.⁵  Activated JAK2 molecules then phosphorylate tyrosine residues in the cytosolic tail of the EpoR that serve as docking sites for signaling molecules, thereby leading to the activation of several intracellular pathways.^6,7  These signaling molecules are often phosphorylated by JAK2 or other kinases such as Lyn.⁸ 

Activation of the EpoR can lead to both antiapoptotic and proliferative effects. For example, signal transducers and activators of transcription 5a and 5b (STAT5a/b) bind to multiple phosphorylated tyrosine motifs (pYs) on the receptor, including pY344 and pY402.^9-11  On tyrosine phosphorylation by JAK2, STAT5 molecules translocate as dimers to the nucleus where they promote transcription of target genes including the antiapoptotic Bcl_xL gene.¹²  Several pathways have been described that independently lead to Epo-induced activation of phosphatidylinositol 3′-kinase (PI3-K) and phosphorylation of its downstream effectors. The p85 subunit of PI3-K directly associates with the most distal tyrosine (Y480) of the EpoR.¹³  Activation of PI3-K can also occur indirectly via Gab adaptors binding to the pY344 and pY402 motifs^14,15  or via phosphorylated insulin receptor substrate-2 (IRS-2), which binds to the EpoR in a nontyrosine-specific manner.¹⁶  Recruitment of PI3-K to the EpoR can lead to a proliferative signal via activation of the mitogen-activated protein kinase (MAPK) pathway. The activation of the MAPK signaling cascade occurs independently of Ras-Raf or STAT5 activation.¹⁷  Administration of PI3-K inhibitors, such as LY294002 to human erythroid progenitor cells, inhibits PI3-K–dependent phosphorylation of Akt and subsequent release of the antiapoptotic Bcl-_xL.¹⁸  Two phospholipase C-γ isoenzymes (PLC-γ1 and PLC-γ2) are direct substrates of tyrosine kinases as they associate with receptors and adaptor molecules via their 2 Src-homology 2 (SH2) domains. Epo-induced tyrosine phosphorylation of PLC-γ2 seems to be regulated by PI3-K^19,20  and requires the presence of pY344, pY402, pY465, or pY480.²¹  There is no evidence for direct interaction of PLC-γ2 with the EpoR, but association with adaptor molecules Gab-2 and Shc has been shown.²⁰  Epo-induced glycosylphosphatidylinositol hydrolysis is correlated with PLC-γ2 activation.^20,21  This leads to the generation of diacylglycerol and inositol-phosphoglycan, which can act as secondary messengers in the mitogenic response.²² 

Several known phosphatases and other inhibitory molecules, involved in negative feedback regulation, associate with different pYs on the EpoR. For example, the protein tyrosine-specific phosphatase SHP-1 binds to the pY344 motif as well as to the JAK2 kinase.^23,24  Other possible inhibitory molecules such as cytokine-inducible SH2-containing protein (CIS) and suppressor of cytokine signaling-2 (SOCS2) are known to associate with pY344 and pY402.²⁵  The interaction of SH2 domain-containing inositol-phosphatase (SHIP) with the EpoR was also demonstrated, but it is not clear whether the association is direct or via the Gab-1 or Shc adaptors.^26,27 

Although many studies report association of signaling molecules to various pYs, it remains controversial what the functional importance of these sites is. Several studies of EpoR mutants, lacking all distal tyrosines, have shown that the membrane-proximal domain of the EpoR is sufficient for receptor functioning. Transgenic as well as knock-in models have shown elevated activity of an EpoR mutant with only the Y344 motif, whereas a diminished activity was observed using a Y-null EpoR.^28-30  Therefore, a negative role for the distal domain of the EpoR has been proposed. Alternatively, in vivo compensatory mechanisms may exist. Li et al demonstrated that the bioactivity of the Y-null EpoR, using physiologic concentrations of Epo, is decreased several-fold and possible compensatory increasing levels of serum growth factors were detected.³¹  The same study also shows that mice expressing the Y-null EpoR were unable to augment erythropoiesis in response to phenylhydrazine-induced anemia. Longmore et al³²  also revealed the requirement of the pYs for efficient red cell development. Together, these reports demonstrate the necessity of the cytosolic tyrosines for optimal EpoR signaling.

To gain more insight into the role of the different tyrosines of the EpoR, we used a recently developed 2-hybrid method in mammalian cells, termed mammalian protein-protein interaction trap or MAPPIT.²⁵  This strategy is based on signaling-deficient receptors with a C-terminal fused bait, and prey proteins coupled to a receptor fragment containing recruitment sites for STAT3. Interaction of bait and prey leads to ligand-dependent activation of STAT3, which can be monitored using a STAT3-responsive reporter gene. In this study, we used pY motifs of the EpoR as baits and evaluated their interaction with CIS, SOCS2, PI3-K, PLC-γ, and STAT5 in different MAPPIT configurations.

Generation of the human EpoR bait constructs in the pcDNA5/FRT vector was as described before.²⁵  Extra leucine residues were inserted in the transmembrane domain of the chimeric receptor using the Quick Change site-directed mutagenesis procedure (Stratagene, Heidelberg, Germany) as described in Lemmens et al.³³  This construct was named pCEL(2L). The motif of the human EpoR (amino acids 318-400) containing only one tyrosine (Y344) was amplified with forward primer 1 (Table 1) containing a SacI site and reverse primer 2 containing a NotI site and stop codon, allowing in-frame coupling to the EpoR-LR-F3 chimeric receptor in pCEL(2L). This construct is further referred to as pCEL(2L)-Y344. Similar bait constructs were made containing 1 or 2 tyrosine motifs of the EpoR: pCEL(2L)-Y432/Y430 (amino acids 404-443), pCEL(2L)-Y444 (amino acids 434-460), pCEL(2L)-Y461/Y465 (amino acids 446-479), and pCEL(2L)-Y480 (amino acids 467-484). The used primer pairs were 3 and 4, 5 and 6, 7 and 8, and 9 and10, respectively. Bait construct Y402 was described previously.²⁵  A Y430F mutation was introduced in pCEL(2L)-Y430/Y432 by polymerase chain reaction (PCR)–based mutagenesis using primers 11 and 12 (Quikchange site-directed mutagenesis method; Stratagene) leading to the pCEL(2L)-F430/Y432 construct. The construct pCEL(2L)-Y430/F432 was also prepared using primers 13 and 14.

The pMG2-CIS plasmid originates from the pMG1-Cis²⁵  and contains aa 905-918 from the gp130 chain in duplicate. The plasmid pMG2-SVT was previously described.³³  Mouse full-length SOCS2 (CIS2) was amplified using primers 15 and 16 on a pEF-FLAG-1/mSOCS-2 plasmid (gift from Dr Starr, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia). After EcoRI-NotI digestion, the fragment was cloned in the pMG2 vector, resulting in pMG2-SOCS2. A part of human PLC-γ2 containing the 2 SH2 domains was amplified from Jurkat mRNA using primers 17 and 18 and was cloned as pMG2-PLCγ(2 × SH2). Mutation of amino acids FLVR to AAAA in the amino terminal SH2 domain of PLC-γ2 with primers 19 and 20 led to the plasmid pMG2-PLCγ(C-SH2) with only one functional SH2 domain. Accordingly, a similar construct, pMG2-PLCγ(N-SH2), was prepared using primers 21 and 22. The plasmid pMG2-p85(2 × SH2) was made by one-step reverse transcription-polymerase chain reaction (RT-PCR; Qiagen, Hilden, Germany) on Jurkat mRNA using primers 23 and 24. Separate amplification of the N-terminal and C-terminal SH2 domain was done with primer pairs 25 and 26, and 27 and 28, respectively, and both were cloned into the pMG2 vector. Full-length mouse STAT5a and STAT5b were amplified with Pfx polymerase (Stratagene) using a RZPD clone (IMAGp998B058512) or the plasmid pECE-STAT5b as template with the primer pairs 29 and 30 or 31 and 32 and ligated in the pMG2 vector. In both constructs a conserved tyrosine in the C-terminal part was mutated using primers 33 and 34 or 35 and 36 to abolish the ability to dimerize and induce promotor activity. These constructs were called pMG2-STAT5a(F694) and pMG2-STAT5b(F699).

The full-length EpoR was amplified using Pfx polymerase (Stratagene) from TF-1 cDNA with forward primer 37 containing a KpnI site and reverse primer 38 containing a stop codon and an XbaI site and cloned in the pSVsport vector by a KpnI-XbaI restriction-based insertion, resulting in the pSV-hEpoR. Different single and double Y→F mutations were introduced by PCR-based mutagenesis using the primer pairs: 39 and 40 for mutation of Y344, 41 and 42 for mutation of Y402, 43 and 44 for mutation of Y430, 45 and 46 for mutation of Y432, 47 and 48 for mutation of Y480 (Quikchange site-directed mutagenesis method; Stratagene) leading to pSV-hEpoR-F344, pSV-hEpoR-F402, pSV-hEpoR-F344/F402, pSV-hEpoR-F430, pSV-hEpoR-F432, pSV-hEpoR-F402/F430, pSV-hEpoR-F402/F432, pSV-hEpoR-F480, and pSV-hEpoR-F432/F480 constructs. A pSV-hEpoR construct with a deletion of amino acids 377-484 was made by mutagenesis with primers 48 and 49 and named pSV-hEpoR-Δ7Y. A second mutation in this construct, Y344F, was done with primers 39 and 40. This EpoR construct was named pSV-hEpoR-F344/Δ7Y. All the pSV-EpoR constructs were transferred into the pcDNA5FRT vector by an XbaI digestion, a Klenow polymerase fill-in reaction, and BssHI digestion.

pcDNA5/FRT vectors expressing the different hEpoR bait constructs were stably integrated in Hek293-Flp-In cells (Invitrogen, Karlsruhe, Germany) as previously described.²⁵  Culture conditions, transfection procedures, and luciferase and β-galactosidase (β-gal) assays for Hek293-T cells are described elsewhere.²⁵  For a typical luciferase experiment, 6 × 10⁵ cells were transfected with the desired constructs in the presence of luciferase and β-gal reporter genes. After 48 hours, cells were left untreated (NS, nonstimulated) or were stimulated with 3.3 ng/mL Epo. After another 24 hours, luciferase activity from triplicate samples was determined and normalized against β-gal activity. Shown results are representative of at least 3 independent transfection experiments. hEpo was obtained from R&D Systems (Minneapolis, MN). The expression of the hEpoR baits was monitored using goat anti–human EpoR polyclonal IgG (R&D Systems) at 2 μg/mL and Alexa 488-conjugated donkey antigoat IgG (Molecular Probes, Eugene, OR) at 4 μg/mL. Fluorescence-activated cell sorting (FACS) analysis was performed on a FACSCalibur (Becton Dickinson, Heidelberg, Germany). Expression of the pcDNA5-hEpoR and hEpoR mutants for relay MAPPIT was verified using the same FACS analysis procedure. Western blot analysis with an anti-FLAG antibody was used to monitor the expression of the different prey constructs using the N-terminal FLAG-tag. The factor-dependent TF-1 cell line was grown in RPMI medium supplemented with 10% fetal calf serum (FCS) and 1 ng/mL granulocyte-macrophage colony-stimulating factor (GM-CSF). After electroporation (300 V, 1500 μF), cells were starved (removal of GM-CSF) for 24 hours and subsequently stimulated with 5 ng/mL hEpo overnight. After 24 hours, luciferase activity was measured as described before.²⁵ 

Hek293-Flp-In cells transiently transfected with the desired constructs were starved for 4 hours in serum-free medium and subsequently stimulated with 5 ng/mL Epo for 15 minutes or were left untreated. Protein concentrations of the nuclear extracts were measured with the Bio-Rad (Hercules, CA) protein assay. Double-stranded oligonucleotides based on the β-casein promoter (sense: CAGATTTCTAGGAATTC; antisense: GGATTTGAATTCC TAGAAATC) were labeled by filling in 5′ protruding ends with Klenow enzyme, using α-³²P]dATP (deoxyadenosine triphosphate; 3000 Ci/mmol; 10 mCi/mL [3.7 × 10⁸ Bq]). This probe binds STAT5 homodimers. Nuclear extracts (15 μg protein) were incubated with about 10 fmol (20 000 cpm) probe in gel-shift incubation buffer (10 mM HEPES [N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid], pH 7.8, 1 mM EDTA [ethylenediaminetetraacetic acid], 5 mM MgCl₂, 5% glycerol, 5 mM dithiothreitol [DTT], 2 mM pefablock SC, 1 mg/mL bovine serum albumin [BSA], 0.1 mg/mL poly (dI-dC)) for 10 minutes at room temperature (RT). The super-shifting STAT5 (Abcam, Cambridge, United Kingdom) or anti–Flag-M2 (Sigma, Munich, Germany) antibodies were incubated with the nuclear extracts for 10 minutes at RT before addition of the radiolabeled β-casein probe. The protein-DNA complexes were separated on a 4.5% (wt/vol) polyacrylamide gel containing 7.5% glycerol in 0.5-fold tris-borate-EDTA (TBE) at 20 V/cm for 90 minutes. Gels were fixed in water/methanol/acetic acid (80:10:10, by volume) for 30 minutes, dried, and autoradiographed.

The MAPPIT system is outlined in Figure 1B and is discussed in detail in Eyckerman et al.²⁵  As basic MAPPIT-bait construct we use a chimeric cytokine receptor composed of the extracellular part of the homodimeric EpoR, fused to the transmembrane and cytosolic domains of the leptin receptor wherein all tyrosines are mutated to phenylalanines (LR-F3). A bait is C-terminally fused to this signaling-deficient receptor. On coexpression with a chimeric gp130-prey construct containing 4 functional STAT3 recruitment sites in the gp130 moiety, complementation can occur, resulting in efficient STAT3 activation. These STAT3 proteins translocate as phosphorylated dimers to the nucleus, allowing direct quantification of STAT3 activation using the STAT3-responsive rat pancreatitis-associated protein 1 (rPAP1) promoter fused to the luciferase reporter gene.²⁵  This approach can be used for analytical purposes, but it can also be adapted to screen complex cDNA libraries. Importantly, the bait-prey interaction may be modification dependent because tyrosine phosphorylation of the bait can occur by the receptor-associated JAKs.

Two extra leucines were inserted in the transmembrane domain to augment the signaling efficiency, as described by Lemmens et al.³³  C-terminal addition of different tyrosine motifs of the cytosolic domain of the human EpoR led to the creation of several EpoR-bait constructs containing 1 or 2 tyrosine residues. These are further referred to as Y344, Y402, Y430/Y432, Y444, Y461/Y465, and Y480 bait (Figure 1B). In case of the Y430/Y432 bait, an additional Y430F or an Y432F mutation was introduced leading to the F430/Y432- and Y430/F432-bait constructs, respectively. Stable expression of all EpoR baits was obtained after Frt recombinase-assisted integration of the different pcDNA5/FRT-EpoR-bait vectors into a Hek-293-Flp-In cell line (Invitrogen). The integration of the EpoR bait occurs in every cell at the same locus so that isogenic cell pools are generated. FACS analysis using an anti-EpoR antibody indicated that the isogenic cell populations showed homogeneous expression of the chimeric EpoR baits. Different expression levels were observed between the different bait constructs (Figure 2 inset).

Several possible interaction partners of the EpoR were cloned as prey. In some cases, both full-length or subfragments containing one or more SH2 domains were made. The preys used in this study are full-length CIS, SOCS2, STAT5a and STAT5b, the N- and C-terminal SH2 domains of PI3-K (p85), PLC-γ, and SHIP. Both STAT5 preys were made signaling deficient by introducing Y694F (STAT5a) or Y699F (STAT5b) mutations. Mutating these tyrosines eliminates binding of the STAT5 prey to the rPAP1-luciferase reporter. As a positive control, we used a prey containing both SH2 domains of PI3-K (p85), which binds the LR-F3 part of the chimeric receptor baits in a pY-independent manner (D. Lavens, T.M., J.T., unpublished results, October 2003). Full-length SV40T was also cloned as a prey and used as a negative control. Expression of all the different preys was checked by Western blot analysis using an anti-FLAG antibody (data not shown).

The different bait-expressing isogenic cell pools were cotransfected with each of the prey expression vectors in combination with the luciferase-based STAT3 reporter (pXP2d2-rPAP1-luci), and interactions were monitored by measuring luciferase activity. Normalized luciferase activities show clear, specific, ligand-dependent responses. No activity was observed in case of expression of an irrelevant SV40T prey, with the exception of the Y344 and Y402 baits, which both showed a very slight activation. The data shown in Figure 2 are corrected for this low background. A prey containing both SH2 domains of PI3-K (p85) was used as a positive control for receptor activation. This prey binds to the LR-F3 cytosolic domain and thus interacts with all chimeric receptor baits. Both SH2 domains seem to be necessary for receptor binding because prey constructs with a single SH2 domain were unable to bind any EpoR-bait construct, except the p85(C-SH2) prey, which shows a slight activation on binding to the Y480 bait (Figure 2 lower right panel).

Strongest luciferase signals for the CIS and SOCS2 preys were obtained through interaction with the Y402 motif. Weaker signals were also observed for the Y344 and Y430/Y432 motifs (Figure 2 upper and middle panels). In the latter case, mutation of Y430 to F in the Y430/Y432 bait leads to reduced binding of the CIS prey, whereas binding of the SOCS2 prey is unaffected (Figure 2 middle panel). Differential binding between these highly related SOCS proteins is also observed for the Y480 motif. Another protein likely involved in negative feedback, SHIP, was also tested for interaction with the EpoR baits. A prey containing the 2 SH2 domains of SHIP shows strongest interaction with the Y344 bait and weaker interactions with the Y402, Y432, and Y480 baits.

A prey containing both SH2 domains of PLC-γ interacts with the Y344, Y402, Y432, and Y480 baits. The separate SH2 domains of PLC-γ were also tested for receptor binding by mutating 4 conserved amino acids (FLVR→AAAA) in either of the 2 SH2 domains of the PLCγ(2 × SH2) prey generating the PLCγ(N-SH2) and PLCγ(C-SH2) preys. MAPPIT experiments showed that receptor binding is to a large extent mediated by the N-terminal SH2 domain of PLC-γ, because mutation of only the N-terminal SH2 domain strongly reduced the luciferase signal (Figure 2).

Both STAT5a and STAT5b are well-characterized partners of the EpoR. To exclude the possibility that STAT5 preys directly induce the luciferase-based reporter gene, the tyrosines Y694 or Y699 were replaced by phenylalanines. These mutated full-length Y694F STAT5a and Y699F STAT5b preys show comparable binding characteristics. Both bind to the Y344 and Y402 motifs (Figure 2 upper panel). Similar experiments wherein both the Y694F STAT5a or Y699F STAT5b prey and an EpoR bait were transiently expressed confirmed the binding of both preys to the Y344 and Y402 bait of the EpoR (data not shown).

The possibility that MAPPIT analysis can detect dynamic, transient interactions was investigated by checking the DNA-binding ability of a STAT5b prey by electrophoretic mobility shift assays (EMSAs) using a probe derived from the β-casein promoter. Nuclear extracts were prepared from Hek293-Flp-In cells transfected with the different bait constructs in combination with either wild-type STAT5b or full-length STAT5b prey. To identify the protein composition of the STAT5b-binding complex, polyclonal antibodies to either STAT5 or to the FLAG-tag were used in a super-shift analysis (Figure 3A-B). Our results indicate that in Epo-treated cells both wild-type STAT5b and the STAT5b prey were able to migrate to the nucleus and bind to the labeled β-casein promoter probe on recruitment to the Y344 or Y402 bait. The other bait constructs were, as expected, not able to activate wild-type STAT5b or the STAT5b prey. Mutation of a single tyrosine (Y699F) in the STAT5b prey completely abolished DNA binding, implying that the observed reporter activity in MAPPIT is exclusively due to activation of STAT3 via the gp130 part of the Y699F STAT5b prey (Figure 3C). These experiments imply that MAPPIT can detect transient bait-prey interactions.

MAPPIT is based on the use of a chimeric receptor wherein a bait is attached to the signaling-deficient cytosolic domain of the LR (LR-F3). In relay MAPPIT, we explored the possibility of relaying the STAT specificity of a wild-type receptor. Because the EpoR signals via STAT5, and not via STAT3, we evaluated whether, in case of interaction with a gp130 prey, a STAT3-dependent signal could be obtained. It is of note that the EpoR uses JAK2, as does the leptin receptor. Clearly, this approach depends on the absence of any detectable rPAP1 reporter activity on activation of the wild-type EpoR. Although the rPAP1 promoter is activated by both STAT3 and STAT5, the very low levels of endogenous STAT5 present in the Hek293-T cells did not lead to any Epo-induced rPAP1 reporter activity (data not shown).

Experiments using the CIS, SOCS2, and PLCγ(2 × SH2) preys showed clear activation of the rPAP1 reporter, supporting this relay MAPPIT concept (Figure 4B). We did not, however, obtain detectable signals using the STAT5a/b or PI3-K p85 preys. By mutating one or more tyrosines in the EpoR, we were able to determine which tyrosine motif in the EpoR is most critical for recruitment of individual prey constructs. Expression of the pcDNA5-hEpoR and hEpoR mutants was monitored by FACS analysis using a goat anti–human EpoR polyclonal IgG (Figure 4A). The data shown in Figure 4B indicate that the Y432 and Y480 motifs are the most critical PLC-γ–docking sites because mutation of one or both tyrosines leads to almost complete loss of the reporter signal. Relay MAPPIT experiments using the PLCγ(N-SH2) and PLCγ(C-SH2) preys indicated that the N-terminal SH2 domain of PLC-γ is preferentially used for the interaction with the wild-type EpoR (Figure 4C). These data are in line with the observations obtained with MAPPIT. Experiments with the basic MAPPIT system showed that CIS and SOCS2 bind to several pYs in the EpoR, although interaction with the Y402 bait gives the strongest signal. Similarly, relay MAPPIT data also show that the Y402 motif is the most prominent interaction site for both (Figure 4B).

We next evaluated the MAPPIT concept in a hematopoietic cell type. Erythroleukemic TF-1 cells electroporated with the pXP2d2-rPAP1-luci reporter showed a clear response with GM-CSF (data not shown). Treatment of TF-1 cells with Epo did not lead to any detectable reporter activity, which is likely due to very low or absent endogenous EpoR expression. We next electroporated TF-1 cells with the collection of EpoR baits in combination with the PLCγ(2 × SH2) prey and the rPAP-luci reporter. The observed prey-binding profile using this transient expression system was very similar to the Hek293-T data set, with prominent binding of PLC-γ to the Y344, Y432, and Y480 motifs. The weaker interaction with Y402 seen in Hek293-T cells could not be confirmed, possibly due to competition by endogenous proteins (Figure 5). No evidence was obtained for other interaction sites.

Although the importance of the EpoR in erythroid development is well established, the precise mechanism as to how the receptor exerts its function is still unclear. Several pathways are activated via the EpoR, including the JAK-STAT, MAPK, PLC-γ, and PI3-K pathways, but the relative importance of each pathway in EpoR signaling is subject to controversy. Over 25 proteins have been described to interact with the EpoR complex including enzymes, adaptors, and various positive and negative regulators. This, and the fact that different pathways can intertwine, make EpoR signaling a complex system. Remarkably, recent data from transgenic as well as knock-in mouse models suggest that the EpoR tyrosine motifs, used for docking of signaling molecules, may be functionally nonessential under optimal circumstances (ie, under non–life-threatening conditions). It is still unknown if compensatory mechanisms, whereby other receptor systems are used, operate in these mice, although elevated levels of serum growth factors were detected.^28-30  Even though mice with an EpoR without any intracellular tyrosine motif are still viable, a recent study using optimized Epo dose-dependent proliferation, survival, and differentiation assays has demonstrated the importance of the tyrosine motifs of the EpoR in abnormal conditions such as phenylhydrazine-induced anemia.³¹ 

A variety of in vivo and in vitro approaches, including activation studies of signaling proteins using truncated or mutated EpoR forms, immunoprecipitation experiments, and phosphopeptide competition assays, have been used to determine the interactions of individual signaling proteins to the different motifs of the EpoR. Most of these techniques study the protein-protein interactions in an indirect manner or in a nonphysiologic environment, sometimes leading to conflicting results. Here, we used a 2-hybrid protein-protein interaction system to evaluate interactions of several signaling proteins with the EpoR in intact human cells.

We could confirm known interactions, such as the binding of STAT5a/b to the pY344 and pY402 motifs.^9,11,34  Previous reports show the phosphorylation and activation of PLC-γ2 on Epo stimulation require the presence of pY344, pY402, pY465, or pY480, but not the recruitment to the EpoR complex.^19-21  We used MAPPIT to demonstrate that PLC-γ2 is recruited to the receptor and associates with different pY motifs of the EpoR. The N-terminal SH2 domain of PLC-γ is preferentially used for these interactions because mutation of only the N-terminal SH2 domain strongly reduces the luciferase signal. Interaction sites of PLC-γ are the pY344, pY402, pY432, and pY480 motifs, whereby the pY432 and the pY480 motifs are most crucial. The PI3-K p85(2 × SH2) prey interacts with the LR-F3 domain, and the precise interaction site is currently mapped in detail (Delphine Lavens, T.M., J.T., unpublished results, October 2003). The strong signal observed in case of the Y480 bait suggests interaction at this site, as has been reported before.¹³  Also, a role for the C-terminal SH2 domain is suggested from our data because activation of this prey occurs at the Y480 motif. These observations require additional analysis using a chimeric receptor bait construct wherein the binding domain in the LR-F3 region is deleted. This will also allow a more detailed analysis of a possible interaction of the PI3-K prey with the Y432 bait. Interestingly, the related proteins CIS and SOCS2 interact with the EpoR in a slightly different manner. Both molecules have the capacity to bind the pY344 motif and are known to bind the pY402 motif on the EpoR.^25,36  Here, we confirm these interactions, but in addition, show binding to the pY430/pY432 motif whereby CIS needs both tyrosines, whereas SOCS2 only needs pY432 for binding to the same motif. Such differential binding is also observed for the pY480 motif.

Comparison of the signal of different preys interacting with a particular motif of the EpoR may give an indication of the binding affinity of a signaling molecule for that motif. Indeed, one could hypothesize that SOCS proteins involved in negative feedback are more likely to bind with high affinity, perhaps irreversibly, to the receptor, whereas signaling molecules, like the STATs, may only bind in a transient manner. In line with this, we observed much stronger signals when SOCS2 and CIS prey molecules are used, compared to STAT5a/b preys, even when these are recruited to the same pY motifs. The transient nature of STAT5b prey binding was confirmed by EMSA, since recruitment of the STAT5b prey leads to STAT3-induced reporter activity, but is also capable by itself to migrate to the nucleus with concomitant promoter binding. It should be stressed that such transient interactions are very difficult to detect using biochemical approaches due to the (affinity) purification process involved. It is also of note that modification-dependent interactions are difficult to analyze in 2-hybrid settings.

We also evaluated a novel MAPPIT-type approach, termed relay MAPPIT. Instead of a signaling-deficient receptor with a fused bait, we used the full-length EpoR itself as a receptor bait, and hence, could analyze interactions under the normal receptor configuration. By mutating one or more tyrosines, we could identify which tyrosine motifs on the EpoR are most critical for interaction with a given prey. With MAPPIT we could show that CIS and SOCS2 can interact with 4 motifs on the EpoR, but using relay MAPPIT we observe most prominent binding with the Y402 motif, since the single Y402F mutation already leads to a strongly reduced signal. This points to the lower sensitivity of this approach. Indeed, weaker signals such as those for the STAT5 preys could not be detected. In case of PLC-γ2, our data suggest that the Y432 and Y480 motifs are the most important binding sites.

The versatility of MAPPIT was further demonstrated by its introduction in hemopoietic cells, for example, the erythroleukemic TF-1 cell line. Electroporation of TF-1 cells with the collection of EpoR baits yielded a data set for the PLCγ(2 × SH2) prey almost similar to what was observed in Hek293-T cells. This confirms that PLC-γ is capable of interacting with the EpoR on several motifs. Similar MAPPIT experiments in TF-1 cells confirmed the relay MAPPIT data showing that CIS and SOCS2 predominantly interact with the EpoR Y402 motif, albeit weaker interactions were confirmed, including the differential binding to the Y430/Y432 motif (data not shown). In addition, we also obtained efficient MAPPIT signaling in BaF/3 cells using βc-based prey constructs (T.M., I.H., J.T., unpublished results, October 2004). These results indicate that the MAPPIT concept is not restricted to the Hek293 cell type, but can be adapted to other cell types, using appropriate STAT-responsive reporter systems.

Together, the findings reported here underline the general robustness of this mammalian 2-hybrid approach. Besides the pY-dependent interactions between cytokine receptors and signaling molecules described here, we recently could also demonstrate other interactions such as those between different TIR domain-containing proteins including Toll-like receptors and their adaptors (P. Ulrichts, J.T., unpublished results, April 2004), between ras and raf (S. Eyckerman, J.T., unpublished results, April 2003) and between F-box proteins involved in cell cycle regulation (M. Caligiuri, personal communication, February 15, 2004). The approach can accommodate different types of modification-dependent interactions because in the heteromeric MAPPIT variant, wherein heteromeric receptor chimeras are used to enforce tethered catalysis, serine phosphorylation-dependent interactions in the transforming growth factor β (TGF-β) signaling cascade could be demonstrated.³³  MAPPIT operates in different types of intact human cells and thus allows analyzing protein-protein interactions in a physiologically optimal environment. Additional characteristics of MAPPIT are its topologic flexibility due to the plasticity of the receptor-bait-prey complex, its ligand-dependent control, and the separation of effector and read-out zones, leading to a reduced background. MAPPIT thus provides an additional tool for the creation of physiologic protein-protein interaction maps.

The authors wish to thank J. Vandekerckhove for continued support and D. Defeau and O. Zwaenepoel for technical assistance.