A complex interaction pattern of CIS and SOCS2 with the leptin receptor.

Hypothalamic leptin receptor signalling plays a central role in weight regulation by controlling fat storage and energy expenditure. In addition, leptin also has direct effects on peripheral cell types involved in regulation of diverse body functions including immune response, bone formation and reproduction. Previous studies have demonstrated the important role of SOCS3 (suppressor of cytokine signalling 3) in leptin physiology. Here, we show that CIS (cytokine-inducible SH2 protein) and SOCS2 can also interact with the leptin receptor. Using MAPPIT (mammalian protein-protein interaction trap), a cytokine receptor-based two-hybrid method operating in intact cells, we show specific binding of CIS with the conserved Y985 and Y1077 motifs in the cytosolic domain of the leptin receptor. SOCS2 only interacts with the Y1077 motif, but with higher binding affinity and can interfere with CIS and STAT5a prey recruitment at this site. Furthermore, although SOCS2 does not associate with Y985 of the leptin receptor, we find that SOCS2 can block interaction of CIS with this position. This unexpected interference can be explained by the direct binding of SOCS2 on the CIS SOCS box, whereby elongin B/C recruitment is crucial to suppress CIS activity.


Introduction
Leptin plays a major role in the regulation of energy homeostasis and food intake. Produced mainly in white adipose tissue (Zhang et al., 1994), it translocates through the blood-brain-barrier to target the leptin receptor (LR) in the hypothalamus. Although six LR splice variants can exist, the LR isoform with an extended cytoplasmic domain (LRlo) is the predominant signalling variant (Ghilardi et al., 1996). A short variant (LRsh) is abundantly expressed in the choroid plexus, brain microvessels, lung and kidney and may participate in leptin transport across the blood-brain barrier (Bjorbaek et al., 1998b;Boado et al., 1998). Next to its effect in weight regulation, leptin is also involved in a broad range of other functions including reproduction, bone formation, growth, immune regulation, angiogenesis and glucose and insulin metabolism.
The LR was addressed to the type I cytokine receptor family based on sequence homology (Tartaglia et al., 1995). It is closely related to the gp130 receptor family, especially gp130, oncostatin M (OSM) and leukaemia inhibitory factor (LIF) receptors, and to the G-CSF receptor (granulocyte-colony stimulating factor) (Zabeau et al., 2003). Leptin typically signals through the JAK-STAT pathway. An overview of LR signalling events is shown in Fig. 1A. The LR carries three conserved tyrosines in its cytoplasmic tail (positions Y985, Y1077 and Y1138 in the murine LR), whereby the membrane distal tyrosine Y1138 is embedded in a STAT3 (signal transducer and activator of transcription) recruitment motif.
The activated receptor recruits STAT3 molecules through their SH2 domain (Baumann et al., 1996;Vaisse et al., 1996), and, after tyrosine phosphorylation, they translocate as homodimers to the nucleus to induce specific gene expression.
Knock-in mice containing a Y1138S mutation reveal a severe obese phenotype but do not show the infertility and reduced size that occurs in db/db mice (Bates et al., 2003). This observation, together with the wide range of leptinresponsive cell types, suggests that alternative signalling pathways must exist. Leptin-dependent activation of STAT1 and STAT5 was demonstrated in vitro (Baumann et al., 1996;Hekerman et al., 2005). In addition, recruitment of SH2containing phosphatase SHP-2 to the phosphorylated Y985 position is responsible for leptin-induced MAPK signalling, although an additional pathway for activation of this signalling cascade directly by JAK2 has been suggested (Bjorbaek et al., 2001). Leptin also induces phosphorylation of IRS-1 and IRS-2 (Duan et al., 2004) and activates phosphatidylinositol 3kinase (PI-3K), as demonstrated in several cell lines (Cohen et al., 1996;Kim et al., 2000). A role for JAK2 in activation of the PI-3K pathway through the JAK2-interacting protein SH2-B and recruitment of IRS-1 or IRS-2 was also reported (Duan et al., 2004). SH2-B and leptin-activated hypothalamic PI-3K both appear essential for weight regulation (Niswender et al., 2001;Ren et al., 2005). Recently, an inhibitory effect of leptin on hypothalamic AMPK (AMP-activated protein kinase) activity was reported. AMPK is proposed to act as a 'fuel gauge' to an intracellular energy sensor cascade and

Hypothalamic leptin receptor signalling plays a central role in weight regulation by controlling fat storage and energy expenditure. In addition, leptin also has direct effects on peripheral cell types involved in regulation of diverse body functions including immune response, bone formation and reproduction. Previous studies have demonstrated the important role of SOCS3 (suppressor of cytokine signalling 3) in leptin physiology. Here, we show that CIS (cytokineinducible SH2 protein) and SOCS2 can also interact with the leptin receptor. Using MAPPIT (mammalian proteinprotein interaction trap), a cytokine receptor-based twohybrid method operating in intact cells, we show specific binding of CIS with the conserved Y985 and Y1077 motifs
in the cytosolic domain of the leptin receptor. SOCS2 only interacts with the Y1077 motif, but with higher binding affinity and can interfere with CIS and STAT5a prey recruitment at this site. Furthermore, although SOCS2 does not associate with Y985 of the leptin receptor, we find that SOCS2 can block interaction of CIS with this position. This unexpected interference can be explained by the direct binding of SOCS2 on the CIS SOCS box, whereby elongin B/C recruitment is crucial to suppress CIS activity.
its activation in the hypothalamus promotes food intake (Minokoshi et al., 2004).
CIS (cytokine-inducible SH2 protein) was the founding member of the SOCS (suppressor of cytokine signalling) family, now consisting of eight proteins: SOCS1-7 and CIS. SOCS proteins typically have an SH2-domain, an N-terminal Overview of LR signalling and its interaction partners. The murine LR carries three conserved tyrosines in its cytoplasmic tail at positions Y985, Y1077 and Y1138. JAK2 is constitutively associated with the LR at the conserved Box 1 and 2 motifs. Upon leptin stimulation, the JAKs become fully activated through cross-phosphorylation and phosphorylate the tyrosine residues in the receptor. STAT3 is recruited to the phosphorylated Y1138 docking site. Upon phosphorylation, STAT3 translocates as dimers to the nucleus, and induces specific gene expression. SHP2 is recruited to the Y985 docking site and couples to the Ras/Raf signalling cascade. The PI-3K pathway is also involved in LR signalling. Tyrosines Y985 and Y1077 take part in negative regulation of the leptin signal by binding SOCS3. PTP-1B is involved in negative regulation by dephosphorylation of JAK2 after internalisation of the LR complex. (B) MAPPIT principle. A particular bait protein is linked C-terminally to the chimeric receptor consisting of the extracellular part of the EpoR and the intracellular part of the LR with all three tyrosines mutated to phenylalanine, whereas the prey protein is fused to the STAT3 recruitment sites of the gp130 chain. The bait-receptor is incapable of recruiting STAT3 upon stimulation. However, when bait and prey proteins interact, the C-terminal part of the gp130 chain is brought in close proximity to the JAK kinases allowing its tyrosine phosphorylation and subsequent STAT3 activation. Read-out is based on a STAT3-responsive reporter construct. (C) GGS-MAPPIT. For GGS-MAPPIT the bait protein is attached C-terminally to a variant of the chimeric EpoR-LR receptor. The cytosolic domain of the LR following the JAK2 association domain is replaced by a GGS-array, preventing any background activation resulting from prey association with the LR-F3. (D) LR-MAPPIT. Here, the LR itself functions as bait protein.
Owing to the Y1138F mutation, no STAT3 recruitment or activation can occur. Upon stimulation, the two membrane proximal tyrosines can nevertheless be phosphorylated by JAK2. Interaction of the prey protein with the LR, which may depend on phosphorylation, allows STAT3 activation and subsequent reporter induction.
preSH2-domain and a C-terminal SOCS box (Starr et al., 1997). The SOCS box targets signalling proteins to the proteasome for degradation by recruitment of an ubiquitintransferase system (Kile et al., 2002). Elongin B or C of the E3 ligase complex is recruited to the BC box in the SOCS box (Kamura et al., 1998). SOCS1 and 3 also carry a KIR (kinase inhibitory region) domain that may act as a pseudosubstrate for direct inhibition of JAK kinase activity. Although SOCS1 associates with JAK2, SOCS3 binds the receptor in close proximity to the kinase and shows only weak affinity for JAK2 (Kubo et al., 2003). Competition for binding to shared recruitment sites can also contribute to the negative regulation of signalling pathways, as exemplified for CIS and SOCS2 in case of STAT5 recruitment at the growth hormone receptor (Greenhalgh et al., 2002a;Ram and Waxman, 1999). SOCS3 was identified as a potent inhibitor of LR signalling. It associates predominantly with the pY985 motif in the LR. Weak interaction at position pY1077 may explain its additive effect on inhibition of LR signalling Eyckerman et al., 2000). SOCS3 is rapidly expressed in the hypothalamus upon leptin stimulation making it part of a STAT3-mediated negative feedback system (Bjorbaek et al., 1998a;Dunn et al., 2005). Recently, PTP-1B was also identified as a negative mediator of LR signalling, targeting both the JAK-STAT and the MAPK pathway (Kaszubska et al., 2002).
It is well established that in many cytokine receptor systems multiple SOCS proteins can be involved in regulation. In the case of the growth hormone, erythropoietin and prolactin receptors, this includes CIS, SOCS2 and SOCS3. Since leptin can activate STAT5 (Baumann et al., 1996;Hekerman et al., 2005) and since CIS and SOCS2 are known regulators of STAT5 recruitment (Ram et al., 1999;Greenhalgh et al., 2002a), we questioned whether CIS or SOCS2 could be involved in LR signalling. Consistent with this, highly conserved tyrosine-based motifs compatible with CIS and SOCS2 association are present in the LR. Also, leptin can induce CIS and SOCS3 expression, and to a lesser extent SOCS2 in insulinoma cells (data not shown). To analyse these interactions with the LR we used two alternative versions of the MAPPIT (mammalian protein-protein interaction trap) strategy ( Fig. 1). We observed differential binding of CIS and SOCS2 with the LR and demonstrate two distinct mechanisms for functional interference by SOCS2.

Cytokine receptor signalling and design of MAPPIT experiments
An overview of signalling through the leptin receptor (LR) is shown in Fig. 1A, and is described in more detail in the introductory section. With MAPPIT we developed a new method to analyse protein interactions in mammalian cells (Eyckerman et al., 2001). MAPPIT bait constructs were originally designed as chimeric receptors, consisting of the extracellular part of the erythropoietin receptor (EpoR) fused to the transmembrane and intracellular regions of a STAT3 recruitment-deficient LR, with a C-terminally attached bait. MAPPIT prey constructs are composed of a prey polypeptide fused to a part of the gp130 chain carrying 4 STAT3 recruitment sites. Co-expression of interacting bait and prey leads to functional complementation of STAT3 activity that can Journal of Cell Science 119 (11) be measured with the STAT3-responsive rat pancreatitisassociated protein I (rPAPI) promoter-luciferase reporter (Fig.  1B). Intrinsic to this strategy, both modification-independent and tyrosine phosphorylation-dependent interactions can be detected.
To monitor interactions with isolated tyrosine motifs of the LR, we developed a MAPPIT configuration whereby the cytosolic domain of the LR is replaced by a large array of Gly-Gly-Ser (GGS) repeats (Fig. 1C). The MAPPIT technique also allows the analysis of interactions with the LR itself by simple mutation of the Y1138 STAT3-recruitment motif to phenylalanine (Fig. 1D). LRs with different combinations of Y to F mutations of the two other conserved tyrosine motifs (located at positions Y985 and Y1077) were used. This allows the study of protein associations with the LR in its normal oligomeric configuration.

MAPPIT analysis of CIS and SOCS2 interactions with the LR
To determine interaction with the LR, the CISprey fusion protein was transiently co-expressed with the LR(YYF) mutant and the luciferase reporter construct ( Fig. 2A). Clear induction of luciferase activity indicated that CIS interacts with the LR. MAPPIT experiments using LR(YFF), LR(FYF) or LR(F3) showed that CIS can interact with both Y985 and Y1077 motifs, whereas no interaction was detected with the LR lacking tyrosines. In a similar way we also tested the SOCS2-LR interaction ( Fig. 2A). SOCS2 clearly associates with the LR, but only at position Y1077. Expression of the LR mutants was analysed using a leptin-SEAP binding assay (Fig. 2B), and expression of the FLAG-tagged CIS and SOCS2 preys was revealed by immunoblotting using an anti-FLAG antibody (Fig. 2C).

Phosphopeptide binding analysis
We confirmed the specific interaction of SOCS2 with pY1077 of the LR using a biochemical strategy (Fig. 3). FLAG-tagged SOCS2 or CIS proteins were expressed in HEK293T cells and total cell lysates were incubated with the biotinylated peptides encompassing the LR phosphorylated or non-phosphorylated Y1077 or Y985 motifs to verify (phospho)tyrosine-specific association. SOCS2 clearly interacted with the phosphorylated Y1077 motif but not with the phosphorylated Y985 motif, confirming its specific phosphorylation-dependent interaction with the LR at position Y1077. Association of CIS was found with neither pY985 nor pY1077 indicating that these interactions may be to weak or short-lived to be detected by phosphopeptide affinity chromatography (data not shown).

Relative binding affinities of the CIS and SOCS2
interactions with the LR To gain further insight into their relative binding affinities for the LR, the CISprey or SOCS2prey were co-expressed with wild-type CIS. Although CIS expression markedly reduced the CISprey signal through both LR(YFF) and LR(FYF), it did not lead to any inhibition of the SOCS2prey signal through Y1077. Conversely, co-expression of wild-type SOCS2 with the CISprey protein clearly diminished the MAPPIT signal at the Y1077 position in the LR whereas the SOCS2prey signal is only partially reduced (Fig. 4A). These results confirm that CIS interactions with Y985 and Y1077 of the LR are weak or transient, whereas the association of SOCS2 with Y1077 is more stable and therefore not easy to compete. It is quite surprising that SOCS2 can inhibit the MAPPIT signal of the CISprey protein through Y985 since SOCS2 is not interacting with this position. Expression levels of the FLAG-tagged proteins were confirmed by immunoblotting using an anti-FLAG antibody (Fig. 4B).

Analysis of CIS and SOCS2 interactions with the LR using GGS-MAPPIT
A new adaptation of the classic MAPPIT method, called GGS-MAPPIT (Fig. 1C), was used to confirm the interaction of CIS and SOCS2 with the LR. In this configuration the cytosolic domain of the LR, following the JAK2 interaction site, is replaced by 60 GGS repeats. GGS triplet repeats are often used as hinge sequences for their known structural flexibility. By using this GGS-MAPPIT strategy any background prey association with the LR-F3 is prevented. The bait constructs containing the LR motifs surrounding Y985 or Y1077 were transiently co-transfected with the prey construct and the rPAP  . SOCS2 interaction with the peptide matching the Y1077 motif of the LR is phosphorylation dependent. FLAG-tagged SOCS2 was expressed in HEK293T cells and lysates were incubated with phosphorylated or non-phosphorylated peptides corresponding to the Y1077 or Y985 motif. Immunoblotting with anti-FLAG antibody revealed specific interaction of SOCS2 with the tyrosinephosphorylated Y1077 motif. luciferase reporter construct in HEK293T cells. Using GGS MAPPIT we were again able to detect the interaction of CIS with both the Y985 and Y1077 motifs, whereas SOCS2 only interacts with the Y1077 motif, but not with the pY985 motif (Fig. 5A). We tested this GGS-MAPPIT strategy further in erythroleukaemic TF-1 cells and obtained similar results as those found in HEK293T cells (Fig. 5B). A full-length FKBP12 bait was used to evaluate non-specific binding of the CIS and SOCS2 preys. FACS analysis, using antibodies against the extracellular domain of the EpoR, allowed monitoring of the expression of the different GGS baits (Fig. 5C).

SOCS2 interferes with STAT5a recruitment
We previously showed that STAT5 can be activated by the LR upon recruitment to the LR Y1077 and Y1138 motifs Journal of Cell Science 119 (11) (Hekerman et al., 2005). Given the strong interaction of SOCS2 at position Y1077 we examined whether SOCS2 can interfere with STAT5 association at this position. The SH2 domain of STAT5a was inserted in a prey construct and used in MAPPIT experiments using the Y1077 motif as bait in GGS-MAPPIT. In HEK293T cells, co-expression of SOCS2 or a SOCS2 mutant lacking the entire SOCS box completely abolished the MAPPIT signal. Similarly, co-expression of SOCS2 ⌬box in TF-1 cells also abrogated the MAPPIT signal, thus excluding a role for elongin B/C recruitment in this suppressive effect (Fig. 6A,B). Similar data were obtained using the LR(FYF) as bait (data not shown). We conclude that SOCS2 can compete with STAT5a association at the pY1077 motif.
SOCS2 interacts with the SOCS box of CIS Given the discrepancy between binding experiments at position Y985, i.e. SOCS2 interferes with CIS binding without interacting with this recruitment motif itself, we examined whether SOCS2 can directly associate with CIS. We first used a MAPPIT configuration with CIS as bait. Here, SOCS2 clearly interacted specifically with full-length CIS (Fig. 7A). In Fig. 7B, we confirmed this interaction by coimmunoprecipitation. Next we looked at association of SOCS2 with the SOCS box of CIS in a MAPPIT experiment and also observed clear interaction (Fig. 7C).
Elongin B/C recruitment is involved in SOCS2 interference with receptor-binding of CIS We developed a mutant of SOCS2, SOCS2(LC-QQ), in which elongin B/C recruitment is abrogated by mutating leucine 163 and cysteine 167, analogous to an elongin B/C recruitmentdeficient SOCS1 mutant reported before (Kamura et al., 1998). We mutated both residues to glutamines to minimise structural alterations. Elongin B/C association was analysed using a twostep purification method, TAP2, based on the classic TAP method (Puig et al., 2001). This sequential purification procedure involves a first protein A tag-based step, followed by TEV protease cleavage to remove the protein A part of the tag and followed by a FLAG-tag-based immunoprecipitation step. Clearly, this SOCS2(LC-QQ) mutant no longer interacted with elongin B or C (Fig. 8A). Furthermore, this SOCS2 mutant as a prey protein still bound CIS in a MAPPIT experiment (Fig. 8B). We next examined the interference of SOCS2 with CIS binding in more detail. At position Y985, the inhibitory effect by co-expression of SOCS2 was completely lost when using the SOCS2(LC-QQ) mutant. Recruitment of elongin B/C to the SOCS box of SOCS2 thus appeared essential for interference with CIS interaction at this position. By contrast, no difference was observed for the SOCS2(LC-QQ) mutant at the Y1077 position, clearly in line with a direct competition with CIS binding at this site (Fig. 8C).

Discussion
MAPPIT allows the study of protein-protein interactions in the physiologically highly relevant context of intact human cells.
Here we used several variations of the MAPPIT concept to study the interactions of two members of the SOCS protein family, CIS and SOCS2, with the murine LR long isoform. CIS and SOCS2 preys were shown to interact with specific tyrosine motifs, either within the full LR configuration or as isolated baits. Interactions were demonstrated in two different cell types: epithelial HEK293T cells as well as the haemopoietic TF-1 cell line.
CIS binding was observed with the conserved mLR Y985 and Y1077 tyrosine-based motifs. By contrast, SOCS2 interacted only at the Y1077 position. In all cases, a Y to F mutation abrogated signalling, indicative of the phosphorylationdependent nature of the interactions. We compared MAPPITbased interaction analysis with a biochemical approach using  HEK293T cells were transiently transfected with the pMET7TAP2-SOCS2 and pMET7TAP2-SOCS2(LC-QQ) constructs. Cell lysates were purified using the TAP2 tag and loaded on a polyacrylamide gel and silverstained. From a parallel experiment, the indicated bands were identified as cullin 5, elongin B and elongin C by mass spectrometry. (B) The SOCS2(LC-QQ) mutant still binds CIS. HEK293T cells were transiently co-transfected with plasmids encoding the chimeric EpoR-LR(F3) construct as a negative control or with the CIS SOCS box bait, and the pMG2-SOCS2 or pMG2-SOCS2 (LC-QQ) prey constructs, combined with the pXP2d2-rPAP1-luci. The transfected cells were either stimulated for 24 hours with Epo or were left untreated (NS, not stimulated). Luciferase measurements were performed in triplicate. Data are expressed as mean fold induction (Epo stimulated/NS) + s.d. (C) Differential effects of the SOCS2(LC-QQ) mutant on CIS interaction with the LR recruitment motifs. HEK293T cells were transiently co-transfected with plasmids encoding different pMet7-LR variants, the pMG2-CIS prey construct, pMet7-FLAG-SOCS2 or pMet7-FLAG-SOCS2(LC-QQ), or the appropriate amount of mock vector together with the pXP2d2-rPAP1-luci. The transfected cells were either stimulated for 24 hours with leptin or were left untreated (NS, not stimulated). Luciferase measurements were performed in triplicate. Data are expressed as mean fold induction (leptin stimulated/NS) + s.d. Leptin receptor and SOCS interactions affinity chromatography with phosphorylated and nonphosphorylated peptides matching the Y1077 or the Y985 motifs. The interaction between SOCS2 and the pY1077 motif was readily demonstrated, in contrast to CIS and its matching phosphopeptides. This is probably due to the more transient or weak nature of the latter interactions. In line with this proposal, competition experiments showed that whereas CIS overexpression could clearly interfere with CISprey binding to either motif, no cross-competition with the SOCS2prey occurred. Conversely, SOCS2 could easily interfere with CISprey binding to pY1077.
Previous reports indicated that the tyrosine at position Y1077 of the receptor was not phosphorylated and was not involved in LR signalling (Banks et al., 2000;Li and Friedman, 1999). However, several observations contradict this supposition. We reported earlier that SOCS3 can interact with the Y1077 domain, although in a rather weak manner, and that this interaction was dependent on tyrosine phosphorylation (Eyckerman et al., 2000). More recently, Y1077 was also reported to induce STAT5 activation (Hekerman et al., 2005). Consistent with a functional role, Y1077 is present in a highly conserved motif, with great similarity to the conserved Y985 domain (Eyckerman et al., 2000). Our findings now lend further support for the important role of the pY1077 motif in LR signalling with two more members of the SOCS protein family interacting at this position, whereby SOCS2 can interfere with CIS and STAT5a prey recruitment.
Very surprisingly, SOCS2 not only interfered with CIS-prey interaction at position Y1077, but also at the Y985 motif without binding this site itself. We provided an explanation for this unexpected finding by showing that SOCS2 directly binds to the SOCS box of CIS. Abrogation of the elongin B/C recruitment ability of SOCS2 had no influence on its association with CIS, but its ability to eliminate CIS receptor binding at position Y985 was completely lost, implying that ubiquitylation and proteasomal degradation of CIS is involved. Very recently, it was reported that SOCS2 also interferes with SOCS3-dependent inhibition of IL-2 and IL-3 signalling (Tannahill et al., 2005). Together, these findings point to an additional, new level of SOCS-mediated signalling control. Reminiscent of this, both mice lacking SOCS2 and SOCS2 transgenic mice exhibit increased growth due to prolonged growth-hormone-dependent STAT5 activity (Greenhalgh et al., 2002b;Metcalf et al., 2000). This dual effect of SOCS2 was also observed in vitro because low SOCS2 doses moderately inhibit GH signalling whereas higher levels positively regulate signalling, probably through interference with SOCS1 function Greenhalgh et al., 2005). Our interaction analysis clearly implicates a complex biological role for SOCS2 and suggests an explanation for the abovementioned duality: SOCS2 can interfere with cytokine signalling through direct interaction with receptors, but can also enhance signalling by eliminating other SOCS proteins through proteasomal degradation. This latter effect may reflect a crucial physiological role of SOCS2 in restoring cellular responsiveness after cytokine activation. In line with this, SOCS2 is usually induced at later time points compared with CIS, SOCS1 and SOCS3 (Adams et al., 1998;Pezet et al., 1999;Tannahill et al., 2005). Detailed quantitative analyses will be required to understand this balancing act in full.
Leptin resistance, which occurs in a majority of obese individuals, may be situated at different levels, e.g. saturation of leptin transport through the blood-brain barrier or aberrations in LR signalling in hypothalamic neurons (El-Haschimi et al., 2000). LR Y1138S knock-in mice are severely obese and fail to activate STAT3, implying a dominant role for STAT3 in leptin-mediated regulation of the energy balance (Bates et al., 2003). Aberrant negative feedback control of LR signalling may contribute to leptin resistance and obesity, because augmented leptin sensitivity and resistance to dietinduced obesity was observed in neural-cell-specific SOCS3 conditional-knockout mice or in SOCS3-haploinsufficient mice (Howard et al., 2004;Mori et al., 2004). In contrast to SOCS3, a negative regulatory role for CIS and SOCS2 on the hypothalamic LR STAT3 pathway is questionable. Bjorbaek and colleagues reported that JAK2 phosphorylation is inhibited by SOCS3 upon leptin stimulation in COS cells but not by SOCS2 or CIS, which both lack a KIR domain at the N-terminus (Bjorbaek et al., 1999). Similarly, we did not observe any clear inhibitory effect on LR signalling through STAT3 by either CIS or SOCS2 (data not shown). This is not unexpected because STAT3 recruitment occurs at the Y1138 motif. Since expression of SOCS2 or CIS is also not upregulated in the hypothalamus upon leptin administration in mice, a role in LR STAT3 signalling is thus unlikely (Bjorbaek et al., 1998a).
CIS and SOCS2 can function through competition with STAT binding at the receptor recruitment site. This mechanism, for example, underlies down-modulation of STAT5 activation by both CIS and SOCS2 upon growth hormone receptor (GHR) activation (Greenhalgh et al., 2002a;Ram and Waxman, 1999). Likewise, the physiological role for CIS and SOCS2 on LR signalling through the Y985 and Y1077 motifs may involve inhibition of recruitment of downstream signalling moieties. This may be particularly relevant in peripheral cell types, known to respond to leptin. Experiments on MLR (mixed-lymphocyte reaction), resulting from the culture of T cells with major histocompatibility complex (MHC)-incompatible stimulator cells, indicated that leptin promotes proliferation of CD4 + T cells (helper T cells, Th) and induces a shifts to activation of Th1 cells, associated with elevated secretion of pro-inflammatory cytokines including interleukin-2 (IL-2) and interferon-␥ (IFN-␥) (Lord et al., 1998). Intriguingly, CIS transgenic mice show altered helper T-cell development with a switch toward Th2 cell response, accompanied by increased IL-4 levels (Matsumoto et al., 1999). CIS may thus be involved in the leptin-dependent modulation of the Th1/Th2 balance. SOCS2 knock-out mice showed a remarkable increase in size whereas growth retardation was observed in CIS transgenic mice. Both SOCS proteins were identified as negative regulators of GHR signalling, presumably through STAT5 (Matsumoto et al., 1999;Metcalf et al., 2000). Considering the decreased linear growth observed in db/db mice and humans with truncated LR (Bates et al., 2003;Clement et al., 1998), SOCS2 and CIS may also exert their influence on growth through regulation of LR signalling. Since LR Y1138S knock-in mice are not reduced in size (Bates et al., 2003) and since, in addition, no clear effect of CIS or SOCS2 was observed on leptin-dependent STAT3 signalling, this effect of SOCS2 and CIS probably occurs independently of STAT3. A good candidate is STAT5, since it is activated in different cell types upon leptin stimulation in vitro through the Y1077 and Y1138 positions in the LR (Baumann et al., 1996;Hekerman et al., 2005). We here showed that SOCS2 can inhibit STAT5a prey association at the LR Y1077 position.
In conclusion, the MAPPIT approach was shown to be a sensitive and flexible system for analysing interactions between proteins in a cellular context. We identified two SOCS proteins, CIS and SOCS2, as new interaction partners of the LR, and identified a novel regulatory role for SOCS2. Full understanding of the biological implications of crossregulation between SOCS proteins on the different cytokine receptor systems will require detailed, quantitative analysis of all members of this protein family.
The pXP2d2-rPAPI-luciferase reporter, originating from the rPAPI (rat pancreatitis-associated protein I) promoter, is used as previously described (Eyckerman et al., 2001). Expression plasmid vectors pEF-FLAG-I/mSOCS2 and pEF-FLAG-I/mCIS were a gift from R. Starr (The Walter and Eliza Hall Institute of Medical Research, Parkville, Australia). Generation of the prey constructs pMG2-CIS and pMG2-SOCS2 both containing part of the gp130 chain (aa 905-918) in duplicate was described elsewhere (Montoye et al., 2005).
The pMET7-FLAG-CIS construct was created by a three-point ligation. The Nterminal FLAG-tag of CIS was cut from the pEF-FLAG-I/mCIS construct using EcoRI-PvuII and the C-terminal part of CIS was cut from the pMG1-CIS construct using KpnI-PvuII as described elsewhere (Eyckerman et al., 2001). Both parts were ligated in the pMet7 vector digested with EcoRI-PvuII. To create the pMET7-Etag-CIS construct, CIS was amplified from the pMET7-FLAG-CIS using primers 5Ј-CGTCCGCGGCCGCGGTCCTCTGCGTACAGGGATC-3Ј and 5Ј-GCTGGCT-CGAGTCAGAGTTGGAAGGGGTACTGTC-3Ј and was ligated in the NotI-XhoI digested pCAGGSE-mMyD88 construct, which was a gift from Rudy Beyaert (Ghent University, Ghent, Belgium). Etag-CIS was then digested with EcoRI-XhoI and ligated in the pMET7 vector.
To generate the pMET7TAP2 construct we used the primers 5Ј-GCGAG-GGCCCGCCACCATGGCCCAGCACGACGAGATCTC-3Ј and 5Ј-CGCTCTCG-AGGCCCTGGAAGTACAGGTTCTCGCTGGCGGTGGTGGGGATGTCGCTGT-TGGCGTCCACGCTG-3Ј to amplify the proteinA domain from the pMA-SpaI construct obtained from the BCCM/LMBP plasmid collection and to add a TEV cleavage site and we cloned the PCR product in the pMET7 vector using ApaI-XhoI. The FLAGtag was introduced using primers 5Ј-GCGAGGGCCCGCCACCAT-GGCCCAGCACGACGAGATCTC-3Ј and 5Ј-GCGAGAATTCCCCGCTGCCCT-TGTCATCGTCGTCCTTGTAGTCCTGGCGCGCGCCCTGGAAGTACAGGTTC-3Ј and cloned in the same vector using ApaI-EcoRI.
pMET7TAP2-SOCS2 was constructed by cutting SOCS2 from the pMG2-SOCS2 construct using EcoRI-NotI and ligating it in the pMET7TAP2 construct.

Cell culture, transfection and reporter assays
Culture conditions, transfection procedures and luciferase assays for HEK293T cells were as previously described (Eyckerman et al., 2000). For a typical luciferase experiment, HEK293T cells were seeded in six-well plates 24 hours before overnight transfection with the desired constructs together with the luciferase reporter gene. Two days after transfection cells were left untreated (not stimulated NS) or were stimulated with 100 ng/ml leptin overnight. The luciferase activity of the transfected cells were measured by chemiluminescence. The factor-dependent TF-1 erythroleukaemia cell line was grown in RPMI medium supplemented with 10% foetal calf serum and 1 ng/ml GM-CSF. After electroporation (300 V, 1500 F), cells were starved (removal of GM-CSF) for 24 hours and were subsequently stimulated with 5 ng/ml hEpo overnight. After 24 hours, luciferase activity was measured as described before.
Leptin binding assay LR expression on the surface of HEK293T cells was measured using a binding assay with a mouse leptin-secreted alkaline phosphatase (SEAP) chimeric protein (Tartaglia et al., 1995). 48 hours after transfection, cells were incubated for 2 hours with a 1:50 dilution of conditioned Cos1 medium containing the chimeric protein with or without an excess of leptin. After two washing steps, cells were lysed in a buffer (1% Triton X-100, 20 mM Tris-HCl pH 7.4) and alkaline phosphatase activity was measured by chemiluminescence, using CSPD substrate (PhosphaLight, Tropix) according to the protocol provided by the manufacturer.

FACS analysis
The expression of the chimeric hEpoR-mLR GGS baits was monitored using goat anti-human EpoR polyclonal IgG (R&D Systems) at 2 g/ml and Alexa Fluor 488conjugated donkey anti-goat IgG (Molecular Probes) at 4 g/ml. Fluorescenceactivated cell sorting (FACS) was performed on a FACSCalibur (Becton Dickinson).

TAP2 purification and mass spectrometry
HEK293T cells were transfected with the appropriate constructs and lysed in cell lysis buffer (50 mM Tris-HCl pH 8, 10% glycerol, 1% NP40, 150 mM NaCl, 5 mM NaF, 5 M ZnCl 2 , 1 mM Na 3 VO 4 , 10 mM EGTA, Complete™ Protease Inhibitor Cocktail). The insoluble fraction was centrifuged and the supernatant was incubated with IgG Sepharose (Amersham Biosciences) overnight. The beads were washed three times with washing buffer (20 mM Tris-HCl pH 7.5, 5% glycerol, 0.1% NP40, 150 mM NaCl) and twice with TEV (Tobacco Etch Virus) protease cleavage buffer 1 (10 mM Tris-HCl pH 8, 150 mM NaCl, 0.1% NP40, 0.5 mM EDTA) and were then incubated with TEV protease in TEV protease cleavage buffer 1 for 2 to 4 hours. The beads were then centrifuged and the supernatant was incubated with anti-FLAG agarose (Sigma) in TEV protease cleavage buffer 2 (10 mM Tris-HCl pH 8, 150 mM NaCl, 0.1% NP40) for 2 to 4 hours. The anti-FLAG agarose beads were washed three times with washing buffer and incubated with FLAG peptide for 10 minutes. The beads were spun down and 4ϫ loading buffer was added to the supernatants before loading on a polyacrylamide gel. Proteins were either visualised by silver staining, or for mass spectrometry analysis with Sypro Ruby protein gel stain according to the manufacturer's instructions (Molecular Probes). Proteins of interest were excised and in-gel digested with trypsin as described. The resulting peptide mixture was dried, re-dissolved in 20 l of 0.1% formic acid in 2/98 (v/v) acetonitrile/water and half of it was applied for nano-LC-MS/MS analysis on an Ultimate (Dionex, Amsterdam, The Netherlands) in-line connected to an Esquire HCT ion trap (Bruker Daltonics, Bremen, Germany). The sample was first trapped on a trapping column (PepMap TM C18 column, 0.3 mm ID ϫ5 mm, Dionex) and after back-flushing, the sample was loaded on a 75 m ID ϫ150 mm reverse-phase column (PepMap TM C18, Dionex). The peptides were eluted with a linear solvent gradient over 50 minutes of 0.1% formic acid in acetonitrile/water (7/3, v/v). Using data-dependent acquisition, only multiple charged ions with intensities above threshold 100,000 were selected for fragmentation. For MS/MS analysis, an MS/MS fragmentation amplitude of 0.7 V and a scan time of 40 milliseconds was used. The fragmentation spectra were converted to Mascot generic files (mgf) using the Automation Engine software (version 3.2, Bruker) and searched using the MASCOT database search engine (http://www.matrixscience.com) against the SwissProt and the NCBInr Database (taxonomy mammalia). Only spectra that exceeded Mascot's threshold score for identify (set at the 95% confidence level) were retained for further manual validation.