Mapping of the leptin binding sites and design of a leptin antagonist.

The leptin/leptin receptor system shows strong similarities to the long-chain cytokine interleukin-6 (IL-6) and granulocyte colony-stimulating factor cytokine/receptor systems. The IL-6 family cytokines interact with their receptors through three different binding sites I-III. The leptin structure was superposed on the crystal structures of several long-chain cytokines, and a series of leptin mutants was generated focusing on binding sites I-III. The effect of the mutations on leptin receptor (LR) signaling and on binding to the membrane proximal cytokine receptor homology domain (CRH2) of the LR was determined. Mutations in binding site I at the C terminus of helix D show a modest effect on signaling and do not affect binding to CRH2. Binding site II is composed of residues at the surface of helices A and C. Mutations in this site impair binding to CRH2 but have only limited effect on signaling. Site III mutations around the N terminus of helix D impair receptor activation without affecting binding to CRH2. We identified an S120A/T121A mutant in binding site III, which lacks any signaling capacity, but which still binds to CRH2 with wild type affinity. This leptin mutant behaves as a potent leptin antagonist both in vitro and in vivo.

Leptin, the product of the ob gene (1), acts as an energy homeostasis hormone. Leptin is secreted into the bloodstream by adipocytes, and blood concentrations of leptin correlate with white adipose tissue mass. Through activation of its receptor in the hypothalamus, leptin can modulate energy expenditure and food intake. Besides this adipostatic function, leptin can also induce proliferation, differentiation, and functional activation of hemopoietic cells (2); it induces angiogenesis (3), enhances wound healing (4), and interacts with the immune and inflammatory responses (5). It enhances T-cell-mediated immune responses by signaling through the long form of the LR on CD4ϩ T lymphocytes (6) and can shift T-cell responses toward a Th1 type, with increased secretion of pro-inflammatory cytokines interleukin-2 and interferon-␥ and decreased interleukin-4 production (7). In experimental mouse disease model systems, the severely obese leptin-deficient (ob/ob) and leptin receptor (LR 1 )-deficient (db/db) mice show reduced experimentally induced colitis, arthritis, and experimental autoimmune encephalomyelitis. Leptin may thus play an important role in the development of autoimmune diseases (8 -12). For recent reviews on the role of leptin in the immune systems, we refer to Matarese et al. (13) and Peelman et al. (14). Leptin also plays a role in atherosclerosis, because leptin promotes many processes of atherogenesis, and ob/ob mice appear to be resistant to diet-induced atherosclerosis (15)(16)(17).
Mature human leptin is secreted as a 146-amino acid protein, with a bundle of 4 helices (helices A-D) with an up-updown-down topology (18). Dali (19) structural similarity searches reveal that leptin shows the highest structural similarity with the cytokines of the IL-6 family and granulocyte colony-stimulating factor (G-CSF) and to a minor extent with other long-chain cytokines, including growth hormone and placental lactogen (20). Similarly, the LR shows highest sequence similarity with the receptors of the IL-6 family and with the G-CSFR. The extracellular part of the human LR contains at least seven structural domains (21). Domains 1 (residue 62-178) and 2 (residue 235-328) have a fibronectin type III fold and together form a cytokine receptor homology module (CRH), named CRH1. Domain 3 (residue 329 -427) has an immunoglobulin (Ig)-like fold. Domains 4 (residue 428 -535) and 5 (residue 536 -635) also have a fibronectin type III fold and together form a second CRH, named CRH2. Domains 6 and 7 adopt a fibronectin type III fold. The presence of an Ig-like domain between two cytokine receptor modules is again similar to the G-CSF and IL-6 family receptors. CRH2 is most likely the main high affinity binding site for leptin on the LR (21). The Ig-like domain is strictly required for JAK2 phosphorylation and concomitant STAT3-dependent signaling (22).
Human IL-6 forms a hexameric 2:2:2 complex with its gp130 and IL-6R␣ chains: each IL-6 molecule binds one IL-6R␣ by its binding site I, and two gp130 molecules by site II and III binding sites ( Fig. 1) (23). Because the leptin/LR system shows significant similarities with the IL-6 family and the G-CSF system, we examined whether similar binding sites I-III are present in leptin. The leptin crystal structure was superposed on the crystal structures of long-chain cytokines, and residues in leptin correlating with binding sites I-III of other cytokines were mutated in mouse leptin. The signaling and binding activity of the different leptin mutants was tested to map the interactions between leptin and its receptor. Interestingly, we identified a leptin mutant in binding site III that lacks any signaling capacity but still binds to the LR with wild type affinity and that thus behaves as a potent leptin antagonist in vitro and in vivo.

Structural Superposition and Molecular Modeling of Mouse Leptin-
The crystal structures of human leptin and other long-chain four-helix bundle cytokines were superposed using the FSSP and Prosup programs. According to this superposition, their sequences were aligned in moe (Chemical Computing Group). A homology model was built for murine leptin by replacing non-identical residues in the human leptin structure by the optimal rotamer of the corresponding residue in mouse leptin, followed by energy minimization, using moe and the charmm22 force field (24). Mutants were chosen based on molecular modeling: solvent-accessible residues selected for mutagenesis were replaced in silico by different hydrophilic residues, and the effect on the overall structure and potential energy of leptin was tested to check mutant misfolding, instability, and hence potential expression problems.
Generation of HA-tagged Mouse Leptin Mutants-The pMET7-SIgK-HA-mLep vector allows the expression of a fusion protein, consisting of the SIgK signal peptide, followed by the HA tag sequence, followed by a four-amino acid GGSG linker, followed by amino acids 3-146 of mouse leptin. Upon expression in eukaryotic cells, the SIgK signal peptide is cleaved off and the HA-tagged protein is secreted in the medium. Mutations in the HA-tagged mouse leptin were introduced using the QuikChange site-directed mutagenesis procedure (Stratagene). Mutations were coupled to a change in restriction cleavage and confirmed by restriction analysis and DNA sequence analysis.
Expression of HA-tagged Mouse Leptin Mutants-3 ϫ 10 6 COS-1 cells were seeded in 75-cm 2 flasks and cultured overnight in DMEM (Invitrogen), supplemented with 10% fetal bovine serum (Invitrogen) in a 10% CO 2 humidified atmosphere at 37°C. The cells were then transfected using a standard polyethyleneimine transfection method with the pMet7-SIgK-HA-mLep vector or mutants in this vector. Medium was replaced 4 h post-transfection, and cells were further cultured for 4 h, after which the medium was replaced by 25 ml of Opti-MEM medium (Invitrogen). After another 90 h, the Opti-MEM medium containing the secreted HA-tagged leptin or mutant was collected, and cells were removed by centrifugation and filtration through a 0.22-m filter. The media were concentrated 30-fold with Vivaspin 15R concentrators, with a molecular weight cut-off of 5000 Da (Sartorius). Expression was verified by Western blot analysis, using a mouse monoclonal anti-HA tag antibody (clone 12CA5, Roche Applied Science), and a peroxidaselabeled goat anti-mouse IgG antibody (Amersham Biosciences). The concentration of the mouse leptin mutants was determined using a mouse leptin enzyme-linked immunosorbent assay kit (R&D systems).
rPAP1-luciferase Reporter Assay-HEK293T cells were transfected with the pMET7-mLRlo plasmid and the pXP2d2-rPAP1 plasmid. The pMET7-mLRlo plasmid encodes the long form of the mouse LR with a C-terminal myc tag. The pXP2d2-rPAP1 plasmid contains the luciferase gene under control of the STAT3-inducible rat pancreatitis-associated protein-1 promoter (25,26). On day 1, 7 ϫ 10 6 HEK293T cells were seeded into a 175-cm 2 flask. Cells were cultured overnight in DMEM (Invitrogen) and 10% fetal bovine serum (Invitrogen) in a 10% CO 2 humidified atmosphere at 37°C. On day 2, the cells were transfected overnight with ϳ36 g of pMET7-mLRlo DNA and 9 g of pXP2d2-rPAP1 DNA using a standard calcium phosphate precipitation procedure. On day 3, cells were washed with phosphate-buffered saline-A (Invitrogen) and cultured overnight. On day 4, cells were detached with cell dissociation agent (Invitrogen) and resuspended in medium, and 50 l of cell suspension was seeded in black 96-well plates (Costar). Cells were stimulated with 50 l of a dilution of the concentrated COS-1 supernatant, containing the HA-tagged leptin mutant. On day 5, the luciferase activity induced by the mutant was measured by chemiluminescence. Cells were lysed in 50 l of lysis buffer ( Competition Assay for CRH2 Binding-We measured binding of the leptin mutants to CRH2 by competition with a mouse leptin-secreted alkaline phosphatase fusion protein (leptin-SEAP, wherein SEAP is fused to the C terminus of leptin) (27). pMET7-mLRCRH2-his6 encodes amino acids 407-604, corresponding to the CRH2 module of the murine LR, followed by a C-terminal His 6 tag. 2 COS-1 cells were transfected with pMET7-mLRCRH2-his6 using the polyethyleneimine transfection method, as described above. Maxisorp plates (Nunc) were coated overnight at 6°C with 0.25 g/ml anti-His5 antibody (Qiagen). Plates were washed four times with wash buffer (phosphate-buffered saline, pH 7.5, containing 0.1% Tween 20). The free protein binding sites on the plates were blocked by incubation with 1% bovine serum albumin (fraction V, Sigma-Aldrich), 5% sucrose in phosphate-buffered saline, pH 7.5, at 37°C for 2 h. Plates were washed four times with wash buffer. The plates were incubated overnight at 6°C with the undiluted medium containing the His-tagged CRH2 of the murine LR. Plates were washed four times with wash buffer. The plates were then co-incubated with diluted COS-1 medium, containing the leptin mutant and a 1/80 dilution of a COS-1conditioned medium containing the leptin-SEAP chimera. After four washing steps with wash buffer, endogenous phosphatases were inactivated (65°C for 30 min), and secreted alkaline phosphatase activity was measured using the CSPD substrate method (PhosphaLight, Tropix) in a TopCount chemiluminescence counter (Packard).
Production and Purification of HA-tagged S120A/T121A Mutant Mouse Leptin-COS-1 cells were seeded at 8 ϫ 10 6 cells per 175-cm 2 flask in DMEM. Ten 175-cm 2 flasks were transfected with the S120A/ T121A pMET7-SIgK-HA-mLep mutant by transfection using polyethyleneimine. After 4 h, the medium was replaced by fresh DMEM, and cells were grown overnight. The medium was then replaced by 50 ml of Opti-MEM, and cells were incubated in this medium for another 72 h. The medium with the secreted S120A/T121A HA-tagged mouse leptin was collected and filtered through a 0.22-m filter, and complete (Roche Applied Science) protease inhibitor was added.
The S120A/T121A HA-tagged mouse leptin was purified onto a 1-ml anti-HA affinity column (Roche Applied Science). The medium was loaded at a flow rate of 0.3 ml/min. The column was washed with 25 ml of equilibration buffer (20 mM Tris/HCl, pH 7.5, 100 mM NaCl, 0.1 mM EDTA) plus 0.05% Tween 20, followed by 10 ml of equilibration buffer without Tween 20. The S120A/T121A HA-mouse leptin was eluted with HA peptide (1 mg/ml) in equilibration buffer at 37°C. The presence of the HA-tagged mutant in the eluates was verified by Western blot analysis, using a mouse monoclonal anti-HA tag antibody (clone 12CA5, Roche Applied Science) and a peroxidase-labeled goat anti-mouse IgG antibody (Amersham Biosciences).
Production and Purification of Recombinant S120A/T121A Mutant Human Leptin-We introduced the S120A/T121A mutation in the pLPPsOMPAhOB plasmid, using QuikChange site-directed mutagenesis as described above. The pLPPsOMPAhOB plasmid harboring the S120A/T121A mutation was electroporated into Escherichia coli MC1061 cells. This allows the expression of untagged, properly folded S120A/T121A human leptin in the periplasm of E. coli MC1061 cells (28). The periplasmic fraction was prepared by an osmotic shock procedure (28), and complete protease inhibitor (Roche Applied Science) was added. The S120A/T121A human leptin was purified from the periplasmic fraction by affinity chromatography with a column consisting of 2A5 monoclonal anti-human leptin antibody (29), coupled to divinylsulfone-activated agarose (Mini-LEAK, Kem-En-Tec). After loading of the periplasmic fraction, the column was washed with phosphate-buffered saline, and the bound S120A/T121A human leptin was eluted with 2 M MgCl 2 . After dialysis against phosphate-buffered saline, residual endotoxins were removed via polymyxin B-agarose (Sigma-Aldrich). The concentration of S120A/T121A human leptin was determined by measuring the optical density at 280 nm. On average, 9 mg of S120A/T121A human leptin was purified from 8 liters of E. coli culture.
Competitive Inhibition of LR Signaling-For testing the antagonistic properties of human and mouse S120A/T121A leptin on mouse LR signaling, HEK293T cells were seeded in 6-well plates, co-transfected with the pMET7-mLRlo and the pXP2d2-rPAP1 plasmids, detached, and resuspended as described above. 50 l of the cell suspension was seeded in each well of a black 96-well plate (Costar). 50 or 100 l of a mixture of mouse or human leptin with mouse or human S120A/T121A leptin was added, as specified below. Cells were lysed after overnight incubation, and the luciferase activity was measured as described above.
For testing the antagonistic properties of the HA-tagged S120A/ T121A mouse leptin, Opti-MEM medium, containing wt HA-tagged mouse leptin was diluted 16-fold in Opti-MEM. 50 l of this diluted medium was added to the cells, together with 50 l of dilution of anti-HA-purified HA-tagged S120A/T121A mouse leptin. For testing the antagonistic properties of the S120A/T121A human leptin, 50 l of cell suspension was incubated with 50 l of a mixture of 9 ng/ml recombinant mouse leptin (R&D Systems) plus different concentrations of purified untagged S120A/T121A human leptin.
For testing the antagonistic properties of human S120A/T121A leptin on human LR signaling, HEK293T cells were seeded in 6-well plates, co-transfected with the pSVSPORT-hLRlo and the pXP2d2-rPAP1 plasmids, detached, and resuspended as described above. The pSVSPORT-hLRlo allows expression of the long form of the human leptin receptor. 50 l of cell suspension was incubated with 50 l of a mixture of 16 ng/ml recombinant human leptin (R&D Systems) plus different concentrations of purified untagged S120A/T121A human leptin.
In Vivo Antagonism of the S120A/T121A Leptin Mutant-DBA/1 mice, 10 -12 weeks of age, were purchased from Janvier (Le Genest St. Isle, France). Mice were treated and used in agreement with the institutional guidelines.
To analyze the effect of WT leptin and leptin antagonist on body weight, either 40 g of WT leptin or 40 g of leptin antagonist were injected intraperitoneally into DBA/1 mice, twice a day. Before injection, both WT leptin and leptin antagonist were preincubated with 200 g of anti-leptin mAb 2A5 to increase their in vivo stability (29). Body weight was measured, three times a week, beginning from 1 day before intraperitoneal injections.

Identification of Candidate Binding Sites in Leptin Using
Structural Superposition-Long-chain four-helical bundle cy-tokines typically interact with their receptor through two or three binding sites in the cytokine. Human IL-6 forms a hexameric 2:2:2 complex with its gp130 and IL-6R␣ chains: each IL-6 molecule binds one IL-6R␣ by its binding site I and two gp130 molecules by its binding sites II and III (Fig. 1A) (23). Binding site I is formed by the C terminus of helix D and binds to the CRH of the IL-6R␣ chain. Binding site II, consisting of residues in helices A and C, interacts with the CRH of gp130. Binding site III consists of residues in the N terminus of helix D, in the loop connecting helices C and D and in the loop connecting helices A and B, and interacts with the immunoglobulin-like domain of gp130. The detailed crystal structure of the Kaposi's sarcoma-associated herpesvirus IL-6 (vIL-6, viral IL-6) in a 2:2 tetrameric complex with the three Nterminal extracellular domains of human gp130 revealed only two binding sites, binding sites II and III, for interaction between vIL-6 and gp130 (30). G-CSF most probably interacts with its receptor in a similar tetrameric complex: binding site II interacts with the CRH of the G-CSFR, whereas binding site III interacts with the immunoglobulin-like domain of the G-CSFR (Fig. 1B) (31). A binding site I is probably not required for G-CSF receptor activation. Site I-III residues were identified in other members of the IL-6 family of cytokines by site-directed mutagenesis: human IL-6, human IL-11, leukemia inhibitory factor (LIF), oncostatin M, and ciliary neurotrophic factor (CNTF) (32)(33)(34)(35)(36)(37)(38).
The mouse leptin model and the crystal structures of human leptin (1ax8), human CNTF (1cnt), human IL-6 (1alu), bovine G-CSF (1bgc), vIL-6 (1i1r), ovine placental lactogen (1f6f), murine LIF (1lki), and human oncostatin M (1evs) were superposed. Binding sites I-III in these cytokines, identified in the crystal structure of the complex with part of their receptor (30,39,40) or by mutagenesis studies (32)(33)(34)(35)(36)(37)(38), were indicated on the structural alignment. Human leptin residues overlapping with or close to these binding sites were considered as possible binding site I, II, or III residues. Within this selection, residues with high solvent accessibility were selected for site-directed mutagenesis. The structural superposition, with identification of the binding sites is shown in Fig. 1E. The alignment corresponding to this superposition is shown in Fig. 1F. Solventaccessible residues outside the three possible binding sites were also selected for mutagenesis to detect possible binding sites that would not coincide with binding sites I, II, or III in other cytokines. These residues were chosen to cover the entire surface of leptin, so that no surface residue is further than 8 Å from a mutated residue. A total of 31 leptin mutants were created. Table I shows a list of the mutations, with their position in the leptin structure indicated.
Effects of Leptin Mutations on LR Signaling-HA-tagged wild type leptin and mutants were expressed in COS-1 cells. Western blot analysis showed expression for all mutants, except for the L13N mutant, which could not be detected (data not shown). The HA-tagged wild type leptin and all 30 expressed mutants had the predicted molecular mass of 16 kDa. Expression levels in the concentrated COS-1 media were determined by enzyme-linked immunosorbent assay and ranged between 5 and 40 g/ml, with the exception of the L13N mutant medium, which showed no detectable leptin expression.
HEK293T cells were co-transfected with a C-terminally myctagged mouse LR and the rat-PAP-luciferase reporter plasmid. Binding of wild type leptin to the LR induces JAK/STAT signaling, with activation of STAT3, which can be detected using the rPAP1-luciferase reporter assay. Fig. 2 shows the luciferase activity induced by the different leptin mutants. HA-tagged wild type leptin shows a sigmoidal dose-response curve up to a maximum value, after which luciferase activity decreases, rather than reaching a stable plateau value. This is also seen in most of the mutants. The data were fitted to a sigmoidal doseresponse curve, and the EC 50 value and maximal luciferase activity were calculated using GraphPad Prism 2.0. The declining part of the dose-response curve following the luciferase activity maximum was discarded in the fitting procedure. Results of these curve fittings are shown in Table I. All the mutant could be fitted with an R 2 of 0.99 or better, with the exception of R20N (0.9607), F41S (0.9872), Q75S (0.9893), and S120A/T121A (0.9192). The R20N and Q75S mutations in the predicted binding site II and the S120A/T121A mutation in the predicted binding site III led to a drastic decrease in luciferase activity, and the corresponding dose-response curves could hardly be fitted to a sigmoidal curve. Remarkably, none of the other mutations led to a drastic increase or decrease in EC 50 values. In contrast, the S29Q/V30Q/S31N, Q34S/R35S, E115S, S117Q, and E122S mutations in predicted binding site III, as well as the F41S and Q138S/Q139S/ V142A mutations in predicted binding site I, led to a decreased maximal luciferase activity. Mutations in predicted binding site II, or outside the three predicted binding sites, did not show a significant effect on the dose-response curves. Mutations with a decreased maximal luciferase activity were mapped in the mouse leptin model structure, as shown in Fig.   3, showing clustering of mutants with a sharply decreased maximal luciferase activity around the inactive S120A/ T121A mutant.
Effects of Leptin Mutations on CRH2 Binding-CRH2 in the LR is thought to form the main high affinity binding site for leptin. We used a competitive binding assay with leptin-SEAP, a fusion protein of mouse leptin coupled to alkaline phosphatase. Binding of the HA-tagged wild type leptin or a mutant thereof to the CRH2 domain inhibits the binding of leptin-SEAP. The competitive binding curves were fitted to a one-site competitive binding curve using GraphPad Prism 2.0. The corresponding IC 50 values for the different mutants are shown in Table II. The D9S/T12Q, K15S, T16N, R20N, Q75S, N82S/D85S, and L86A mutants in the predicted binding site II display IC 50 values that are at least 20 times higher than the IC 50 value of the HA-tagged wild type leptin. Mutations that affect residues outside the predicted binding site II did not lead to a similar striking increase of IC 50 value but showed a similar competitive binding behavior as the HA-tagged wild type leptin. Even the S120A/T121A mutation in predicted binding site III, which inhibits JAK/STAT sig- naling, did not lead to an increased IC 50 value. Mutations with a decreased CRH2 binding activity are all found in helices A and C in the mouse leptin model structure, as shown in Fig. 4.
The S120A/T121A Leptin Mutant Behaves as an Antagonist in Vitro-The HA-tagged S120A/T121A mouse leptin mutant avidly binds to the CRH2 domain of the LR but fails to induce proper STAT3 activation. This suggests that this mutant is able to bind to the receptor without activating it. Moreover, we could show that it is able to inhibit leptin-SEAP binding to the CRH2 domain, with an IC 50 value comparable to that of wild type leptin. We therefore tested whether the S120A/T121A leptin mutant could act as a competitive inhibitor of LR activation. HA-tagged S120A/T121A mouse leptin and S120A/ T121A human leptin were purified as described under "Experimental Procedures." No contaminants could be detected by SDS-PAGE with silver staining when 10 g of protein was analyzed (data not shown). HEK293T cells were co-transfected with a C-terminally myc-tagged mouse LR and the rat-PAPluciferase reporter plasmid. These cells were incubated with HA-tagged mouse leptin plus different dilutions of purified HA-tagged S120A/T121A mouse leptin. To test the effect of the untagged S120A/T121A human leptin, cells were incubated with recombinant mouse leptin plus different concentrations of purified S120A/T121A human leptin. In both cases, luciferase reporter activity was determined. Both the untagged S120A/ T121A human leptin (Fig. 5A) and the HA-tagged S120AT121 mouse leptin (data not shown) inhibited the mouse LR signal-ing in a dose-dependent manner. Therefore, both acted as potent antagonists of the mouse LR. The antagonist showed species specificity: the human S120A/T121A leptin mutant was a less potent antagonist for mouse LR signaling: about 10-fold lower concentrations of untagged S120A/T121A human leptin were needed to completely inhibit human LR signaling (Fig. 5B).
In Vivo Antagonism of the S120A/T121A Leptin Mutant-Because leptin was able to regulate the appetite and energy expenditure, we assessed whether in vivo administration of the S120A/T121A leptin mutant had an antagonistic effect on the body weight. Therefore, male DBA/1 mice were injected intraperitoneally with either phosphate-buffered saline, 40 g of wild type leptin, or 40 g of S120A/T121A leptin mutant, twice a day. Before injection, both WT leptin and leptin antagonist were preincubated with 200 g of anti-human leptin mAb 2A5. Co-administration of 2A5 monoclonal antibody increases the half-life of leptin or leptin antagonist in circulation (29). Administration of 2A5 monoclonal antibody alone did not have a significant effect on body weight. Fig. 6 shows that in vivo administration of the S120A/T121A leptin mutant resulted in increased body weights, which were significantly different from the 2A5-treated group by day 8 (p Ͻ 0.03, Mann-Whitney test). Treatment of mice with exogenous wild type leptin resulted in declined body weights, significantly different from the 2A5-injected group by day 8 (p Ͻ 0.01, Mann-Whitney test).

DISCUSSION
Because the leptin/LR system is related to the G-CSF and the IL-6 family of cytokines/receptors, we compared the structure of leptin with the structures of these long-chain four-helix bundle cytokines. In analogy with these systems, three possible binding sites I-III were identified in leptin, and residues in these binding sites were mutated. To exclude the existence of additional binding sites in other parts of the leptin surface, mutations were also made in other parts of the surface of leptin. The effect of the mutations was analyzed on LR activation using a reporter-based assay, and on binding to CRH2, which is considered to be the major leptin binding domain in the LR (21).
The F41S and Q138S/Q139S/V142A mutations in predicted binding site I led to a decreased maximal luciferase activity. However, even the Q138S/Q139S/V142A mutation, which affects three residues predicted to be in the center of a possible binding site I, did not lead to complete loss of activity. This suggests that this binding site is not strictly required for receptor activation. In contrast, with binding sites II and III, the residues in binding site I in leptin were not conserved between different species. In human leptin, for example, glutamine 139 was replaced by a tryptophan right in the center of predicted binding site I. It is very well possible that human and mouse leptin differ in their behavior, when it comes to the usage or importance of binding site I. In the gp130 family of cytokines, binding site I binds to the CRH domain of a non-signaling ␣-receptor chain. In analogy, CRH1 or CRH2 of the LR may form a binding site I interaction site.
In analogy with the other four-helix bundle cytokines, a major binding site could be expected at the surface of the antiparallel helices A and C. In these cytokines, this binding site II interacts with a CRH domain. We designed five mutations at the surface of helix A (D9S/T12Q, L13N, K15S, T16N, and R20N) and six mutations at the surface of helix C (Q75S, N78S, N82S/D85S, L86A, L89A, and F92A). The R20N and Q75S mutations had a drastic effect both in the reporter assay, and on CRH2 binding, which is in agreement with the inhibitory effects of the R20Q mutation reported by Verploegen et al. (29). These residues are on adjacent positions in helices A and C and interact with each other (Fig. 4). An effect of these mutations on the overall protein structure cannot be excluded. None of the other mutations led to a clear effect on EC 50 values in the reporter assay, although the values for maximal luciferase activity seem to be slightly decreased. In contrast, the D9S/T12Q, K15S, T16N, N82S/D85S, and L86A mutations showed a strongly decreased binding affinity for the isolated CRH2 domain. Our data confirm the existence of a binding site II in leptin, which interacts with CRH2 in the LR.
Fong et al. (21) and Zabeau et al. (22) showed that CRH2 is critical for leptin binding. When tested in surface plasmon resonance assays, a recombinant CRH2 domain, expressed in Escherichia coli shows a similar affinity as the entire extracellular domain of the LR, expressed in Sf9 cells (15.3 versus 9.5 nM) (41,42). Sandowski et al. (41) suggested that other parts of the extracellular domain might play only a minor, if any, role in leptin binding. This seems to be in sharp contrast with our data, which show that mutations with a sharply decreased affinity for CRH2 have no effects on the EC 50 value for LR activation. The apparent discrepancy might be explained by assuming that the LR exists as a preformed complex, in which the actual binding site is formed by binding epitopes of multiple LR chains. In such a case, mutations that inhibit binding to CRH2 might still bind to the more extensive composite binding site. Several observations indicate that the LR indeed exists as a preformed complex (43)(44)(45)(46).
Mutagenesis studies of different members of the gp130 family of cytokines revealed the existence of a unique binding site  III in these cytokines (32)(33)(34)(35)(36)(37)(38). This binding site is a discontinuous site formed by the residues at the N terminus of helix D (site IIIa) and by residues in the AB loop (site IIIb). The crystal structure of the Kaposi's sarcoma-associated herpesvirus IL-6 (vIL-6, viral IL-6) complex with the three N-terminal extracellular domains of human gp130 showed that binding site III in vIL-6 interacts with the Ig-like domain of gp130 (30). A similar interaction was shown between binding site III in G-CSF and the Ig-like domain of its receptor (31). Likewise, binding site III of leukemia inhibitory factor (LIF), cardiotropin-1, and oncostatin M all interact with the Ig-like domain of the LIF receptor (47). We tested the existence of a binding site III in leptin by mutagenesis of residues at the N terminus of helix D and the preceding CD loop and of residues in the AB loop. In the reporter assay, the S29Q/V30Q/S31N and Q34S/R35S mutations in the AB loop, and the E115S, S117Q, and E122S mutations in the CD loop led to a decreased maximal luciferase activity, without altering the EC 50 value. Their binding to CRH2 was unaltered. The most drastic effect was seen in the S120A/T121A mutant, which had totally lost the ability to activate the receptor but was still normally bound to CRH2. Ser-120 and Thr-121 are found exactly at the Nterminal of the D helix. In gp130 cytokines, binding site III is mainly determined by a hydrophobic residue and an adjacent basic residue at this location, and mutagenesis of these residues equally leads to an inactive cytokine. The most likely interaction site for the leptin binding site III is the Ig-like domain of the LR. In vitro binding experiments show that leptin has no appreciable affinity for an isolated recombinant CRH1-Ig protein. 2 In our present work, none of the mutations outside the three predicted binding sites had a clear effect on the STAT3induced luciferase activity or on the binding to CRH2. Leptin thus seems to behave like the gp130 family of cytokines, with binding sites II and III, and possibly a binding site I. Leptin binds to soluble forms of the extracellular part of the LR in a 2:2 ratio (44,45). This complex might resemble the 2:2 G-CSF⅐G-CSF receptor complex and might be formed by interactions of binding sites II and III of leptin with CRH2 and the Ig domain (Fig. 1, B and D). Binding site I might be involved in the binding of additional LR chains (Fig. 1C), thus leading to higher order complexes, possibly with enhanced signaling. Such a signal-enhancing role for binding site I FIG. 5. Antagonistic properties of S120A/T121A human and mouse leptin in vitro. A, antagonistic effect of the recombinant S120A/T121A human leptin on mouse LR signaling. HEK293T cells were transfected with the pSVSPORT-hLRlo and pXP2d2-rPAP1 plasmids and incubated overnight with 9 ng/ml recombinant mouse leptin (R&D Systems) plus different concentrations of purified untagged S120A/T121A human leptin. Luciferase reporter activity (CPS: counts per second) is plotted as a function of the antagonist concentration. B, antagonistic effect of the recombinant S120A/T121A human leptin on human LR signaling. HEK293T cells were transfected with the pSVSPORT-hLRlo and pXP2d2-rPAP1 plasmids and incubated overnight with different concentrations of wild type human leptin (diamonds), with different concentrations of S120A/T121A human leptin (triangles), or with 16 ng/ml recombinant human leptin plus different concentrations of purified untagged S120A/T121A human leptin (crosses). Luciferase reporter activity is plotted as a function of the leptin/antagonist concentration. would resemble the signaling behavior of Kaposi's sarcomaassociated herpesvirus IL-6. This viral IL-6 is able to activate the gp130 receptor without using its binding site I. However, recruitment of IL-6␣ receptor chains through binding site I of viral IL-6 increases the signal (48). Using a JAK/STAT complementation assay, we were able to show that activation of the LR can happen through a higher order complex, with more than two receptor chains per complex (22).
Receptor activation probably requires initial binding of leptin to a composite binding site formed by the LR dimer or oligomer. The interaction of the different binding sites in the leptin molecule with specific domains of the LR brings the receptor chains in the proper orientation for receptor activation. This hypothesis explains why mutations often have more effects on the maximal receptor activation than on the EC 50 value: the EC 50 value is determined by the affinity of leptin for the receptor and is comparable to the K d of leptin for the LR (21,41,42,49). Due to the composite binding site on the LR, most mutations have only little influence on the affinity for the total receptor complex and on the EC 50 value. Receptor activation, however, requires the interaction of a specific binding site in the leptin molecule with a specific domain in the LR, and this is much more affected by the point mutations. Therefore, many mutants are less able to activate the receptor, leading to a lowered maximal reporter activity.
The S120A/T121A mutation shows normal binding to CRH2, but is completely unable to activate the LR. As expected, the S120A/T121A leptin mutant (both human and mouse) behaves as an antagonist and blocks activation of the LR in a dose-dependent manner. This antagonistic action of S120A/T121A is reminiscent of the antagonistic effect of IL-6 and IL-11 binding site III mutants (50 -52).
Another leptin antagonist R128Q has been described (29,53). R128Q is not part of any of the three predicted binding sites and is largely buried in the crystal structure, where it forms hydrogen bonds with the backbone of proline 39 and valine 109. The R128Q mutation probably disturbs the proper orientation of the AB and CD loops, and thus possibly indirectly affects binding sites I and III. We compared the antagonistic properties of the R128Q and the S120A/T121A mutant.
Although the R128Q mutant still shows LR activation at higher concentrations, this is not seen with the S120A/T121A mutant (data not shown).
In this report we have examined the in vivo effect of the S120A/T121A leptin mutant. We demonstrated that the body weights of mice are increased following daily injection of the S120A/T121A leptin mutant, indicating the S120A/T121A leptin antagonist stimulates feeding and/or reduces energy expenditure. Antagonizing leptin has been suggested as a possible therapy in auto-immune diseases (11,54) and might also have beneficiary effects on atherosclerosis. The S120A/T121A leptin antagonist offers a novel tool to delineate the precise role of leptin in human disease.