468699.qxd 7/31/03 10:28 AM Page 335
Journal of Protein Chemistry, Vol. 22, No. 4, May 2003 (© 2003)
Modeling and Docking of the Three-Dimensional Structure of the Human Melanocortin 4 Receptor Xiaonan Yang,1,2 Zhuorui Wang,2 Wei Dong,2 Lunjiang Ling,1 Huanming Yang,2 and Runsheng Chen1,2,3 Received December 2, 2002
A three-dimensional structure of the human melanocortin 4 receptor (hMC4R) is constructed in this study using a computer-aided molecular modeling approach. Human melanocortin 4 receptor is a G Protein-Coupled Receptor (GPCR). We structurally aligned transmembrane helices with bovine rhodopsin transmembrane domains, simulated both intracellular and extracellular loop domains on homologous loop regions in other proteins of known 3D structure and modeled the C terminus on the corresponding part of bovine rhodopsin. Then tandem minimization and dynamics calculations were run to refine the crude structure. The simulative model was tested by docking with a triplet peptide (RFF) ligand. It was found that the ligand is located among transmembrane regions TM3, TM4, TM5, and TM6 of hMC4R. In consistence with mutational and biochemical data, binding site is mainly formed as a hydrophobic and negatively charged pocket. The model constructed here might provide a structural framework for making rational predictions in relevant fields. KEY WORDS: hMC4R; GPCR; homology modeling; docking.
1. INTRODUCTION
found to express only in the brain and be involved in the Leptin pathway; it contributes to the controlling of human feeding behavior and energy homeostasis (Barsh et al., 2000; Schwartz et al., 2000). To our better knowledge, hMC4R gene’s 2 different frameshifts and several mutations have been identified associating with human obesity (Farooqi et al., 2000; Vaisse et al., 1998; Yeo et al., 1998). Relevant natural agonists and antagonists of MC4R were also discovered, including -MSH and AGRP (Ollmann et al., 1997; Schioth et al., 1996). A triplet peptide: Arg-Phe-Phe (RFF) was then testified as the smallest conserverd motif of AGRP which mediates key interactions between its analogs and MC4R (Tota et al., 1999). GPCRs, membrane-bound proteins, have seven transmembrane helices and are of special importance due to their forming of the largest and the most diverse receptor family; they participated in a lot of cell signal transduction processes. In general, GPCRs are activated by binding of extracellular, receptor-specific ligand through extracellular or transmembrane domains. Such activations can change GPCRs conformations and then activate the intracellular heterotimeric G proteins which
The human melanocortin 4 receptor, one member of melanocortin receptor family, is a 332-amino acid GPCR (Gantz et al., 1993), and melanocortin receptors (MCR) mediate signal transduction induced by melanocortin peptide via G-protein coupled signaling system. In human, five receptor isotypes that share high homology in primary amino acid sequences have been identified (MC1R-MC5R), but they differ in terms of their physiological roles as skin pigmentation, anti-inflammatory effects, regulation of blood pressure, immune modulation and food intake (Chhajlani and Wikberg, 1992; Chhajlani et al., 1993; Gantz et al., 1993, 1994; Mountjoy et al., 1992; Roselli-Rehfuss et al., 1993). Human MC4R is
1
2
3
Laboratory of Bioinformatics, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China. Beijing Genomics Institute/Center of Genomics and Bioinformatics, Chinese Academy of Sciences, Beijing, China. To whom correspondence should be addressed at Laboratory of Bioinformatics, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, China. E-mail:
[email protected]
335 0277-8033/03/0500-0335/0 © 2003 Plenum Publishing Corporation
468699.qxd 7/31/03 10:28 AM Page 336
336 sequentially stimulate several kinds of signaling cascade such as cAMP and PKA pathways (Gether and Kobilka, 1998; Ji et al., 1998; Lefkowitz, 1998). Two models of human melancortin 1 receptor (hMC1R), which is 44% homology with hMC4R, have been reported (Prusis et al., 1995, 1997). The researchers used homology modeling method but based them respectively on bacteriorhodopsin and 9Å resolution model of bovine rhodopsin (a prototypical G protein-coupled receptor). Presently, the first high-resolution (2.8Å) 3D structure of bovine rhodopsin has been acquired via X-ray crystallography (Palczewski et al., 2000). This structure made further refinement by offering a new opportunity to assemble the simulation studies with other GPCRs. In this study, we constructed a three-dimensional model of the human melanocortin 4 receptor by basing its structure on the 2.8Å structure of bovine rhodopsin to which it has sequence homology. Since the melanocortin 4 receptor is linked to G-protein, it is important to understand how binding of ligands to this receptor induces
Yang, Wang, Dong, Ling, Yang, and Chen activation of the G-protein in a signal transduction cascade. We also provided the results of ligand binding that might be applicable for rational drug design. To the best of our knowledge, this work should be the first study of modeling human MC4R 3D structure using the rhodopsin crystal structure rather than the electron density projections structure.
2. METHODS 2.1. Modeling the 3D Structure of the hMC4R The strategy for constructing the model of hMC4R based both on structural alignments with rhodopsin and use of secondary structure prediction algorithms. Crude receptor structures were obtained from a graphic modelbuilding technique with the aid of SGI graphic workstation and software packages: InsightII, Discover3, Homology, Biopolymer and Docking of MSI Inc. (Chen et al., 1996).
Fig. 1. Two-dimensional model of hMC4R (without the 1-41 residues of N terminus). Each darkened helical column represents of the residues defining the seven TM helical regions. The conserved residues, which have been restrained during the energy minimization procedure, are show in a dark and heavy style. Other two horizontal helical bars, which represent the -helical regions of I3 and C terminus, are show in gray circles.
468699.qxd 7/31/03 10:28 AM Page 337
Structure of the Human Melanocortin 4 Receptor 2.1.1. Sequence Alignment To discriminate TMs location of hMC4R, multiple sequence alignment was undertaken using CLUSTAL W 1.82, by aligned sequences of 24 melanocortin receptors and 25 rhodopsins. The Blosum scoring matrix was selected with the penalty value for opening a gap setting at 10, the gap ends was set at 10, and the penalty for extending a gap and separation of a gap was set at 0.05. Sequence of hMC4R, all other different species and subtypes of the MCRs and rhodopsin families were found in EMBL database.
2.1.2. Structural Alignment Once the initial position of the transmembrane domain was achieved in the sequence of hMC4R, the sequence of each helix was structurally aligned with that of corresponding helix of bovine rhodopsin by Homology program. Amino acid residues of bovine rhodopsin (PDB
337 ID: 1HZX; Teller et al., 2001) were mutated to those of hMC4R to construct the initial structure of the helices of this receptor. To construct intracellular and extracellular loop domains, in addition with the results of Blastp, an original loop list from the InsightII (ftp://ftp.msi.com/pub/dist/ pdb95_ca_dist.list) was used to find the fitted loop templates using Biopolymer program. Secondary structure predictions had been carried out before this stage, using at least four different methods. The prediction previews were particularly considered in the template selection process. The secondary structure prediction were samely implemented and referred to before the C terminus’s template selection. After selecting the templates for both loop and C terminus domains, the appropriate side chains in the selected templates were modified through interactive residue substitution to setup the loop and terminus crude structure.
Fig. 2. Sequence alignment of the TM and Loop domains of hMC4R. (a) Sequence alignment of the seven transmembrane (TM) segments of MC4R with bovine rhodopsin (1HZX). In all sequence alignment figures, an asterisk (*) indicates an identical amino acid, “:” indicates a conserved amino acid, “.” indicates a half conserved amino acid. (b) Sequence alignments of loops segments of MC4R with their corresponding templates. In all sequence alignment figures, an asterisk (*) indicates an identical amino acid, “:” indicates a conserved amino acid, “.” indicates a half conserved amino acid.
468699.qxd 7/31/03 10:28 AM Page 338
338 The last 12 residues of the hMC4R and the N terminus (1-41) did not model in this process because of little information obtained for their precise extracellular locating.
Yang, Wang, Dong, Ling, Yang, and Chen at all. The TM domains were relaxed gradually during this simulation procedure.
3. RESULTS AND DISCUSSION 2.1.3. Refinement of the hMC4R Then we linked the TMs loops and C terminus of the target molecule, and apply energy minimization to refine the flank two residues of the new peptide bond to 0.01 kcal/mol·Å. Discover3 program was used to do the energy minimization and dynamics simulation process (Ling et al., 1989). Hydrogen atoms were added and the crude model was energy minimized under CVFF forcefield with several conserved residues of each TM domain restrained contemporary (Fig. 1). The Van der Waals cut off was set at 12Å, and the other no bond cut off at 14Å. Then, 4000 steps steepest descent was done. The conjugate procedure was then performed when the rms value reached 0.1 kcal/mol·Å. Following that, molecular dynamics calculation was executed on the extrcellular and intracellular loops: 1000fs promoted to 298 K, 298 K was equilibrated at for 5 ps, and 50 ps simulation was set to find the low conformation of hMC4R model. The step was 1fs. The selected lowest conformation was finally energy-minimized. During this minimization procedure, the restraints put on the transmembrane domains were reduced gradually.
3.1. hMC4R 3D Structure 3.1.1. Sequence Alignment and Structural Alignment Results An outline of the TMs and Loops locations is shown in Fig. 1. Multiple alignment result showed several conserved residues in each TM domain both of MCR and Rhodopsin familys which are consistent with several former researches (Baldwin et al., 1997; Prusis et al., 1997). Assuming that the membrane thickness is usually the same, hMC4R TM domains were defined as equal number residues as bovine rhodopsin in accordance with the multiple alignment result. There was only one exception
2.2. Ligand-Receptor Complex Modeling Triplet peptide-RFF, the key motif of natural antagonist of AGRP, was selected to do the next docking research. We extract its coordinate from the NMR structure of AGRP (PDB ID: 1HYK) (McNulty et al., 2001). 2.2.1. Docking the Ligand with MC4R There were two key stages in the docking procedure. According to the hydrophobic and electric charge surface analysis, the first stage was manually putting the triplet peptide into a putative docking pocket. It must be reminded that avoiding the steric overlap between ligand and receptor during this stage was required. Then dock using Grid/Monte carlo Minimization method (Zhu et al., 2002). The grid resolution is 0.5Å. A distancedependent dielectric constant was used. 2.2.2. Refinement of the Ligand-Receptor Complex After docking, energy minimization was applied on the whole complex. No restraints were put on the ligand
Fig. 3. Three-dimensional model of melanocortin-4 receptor. The modeled hMC4R draw as a ribbon diagram with InsightII.
468699.qxd 7/31/03 10:28 AM Page 339
Structure of the Human Melanocortin 4 Receptor that the TM5 domain was shorted ~2 residues because if not, there would be only 2 residues left for extracellular loop2. Analysis of the TM regions between hMC4R and the template molecule shows a significant identity for each pair (Fig. 2a). The percentage identities (and similarity-identical plus conservative substitutions) for each of the TMs between hMC4R and rhodopsin are: TM1: 13.3% (43.3%), TM2: 26.7% (60%), TM3: 18.2% (42.4%), TM4: 4.3% (34.8%), TM5: 25% (50%), TM6: 16.1% (58.1%), TM7: 14.3% (52.4%). The adopted templates for each of the loops were shown in Fig. 2b. Besides consulting the secondary structure prediction results, three extra criteria were consulted during the loops’ template selection: (1) the template structure had to be between two anti-parallel -helix; (2) the length for the candidate loop must be close to the target loop; (3) the sequence identity of the template protein with the target protein had to be ≥25%; if not, the
339 distance between the ends of loop in the template structure need to differ by no more than 2Å from the distance between the TMs to be connected in hMC4R. Based on the secondary structure prediction results, we selected the corresponding part of bovine rhdopsin’s C-terminal as the template for hMC4R C terminus. The prediction result is consistent with the template structure owned an -helical structure. And the residues number is equal too. Furthermore, they both ended by two cystins that were conserved in many rhodopsin family GPCRs (Palczewski et al., 2000).
3.1.2. Dynamics and Energy Minimization Simulation Results After initial energy minimization, total energy of MC4R reduced to 0.34 × 104 kcal/mol from a very high
Fig. 4. The hydrophobic surface and electric charge surface of putative binding pocket. (a) Surface representation of the hydrophobic surface. Red to blue shading indicates residues with decreasing hydrophobic shift. (b) Surface representation of the electric charge surface. Red to blue shading indicates residues with increasing negative to positive charge shift.
468699.qxd 7/31/03 10:28 AM Page 340
340
Yang, Wang, Dong, Ling, Yang, and Chen
Fig. 4. Continued. The hydrophobic surface and electric charge surface of putative binding pocket. (a) Surface representation of the hydrophobic surface. Red to blue shading indicates residues with decreasing hydrophobic shift. (b) Surface representation of the electric charge surface. Red to blue shading indicates residues with increasing negative to positive charge shift.
potential energy level. In the next molecular dynamics simulation procedure, the crude structure of hMC4R was simulated for 50 ps. The lowest potential energy 3D structure of the last-10 ps simulation (denoted MC4Rlow) was selected for the next minimization procedure. The MC4Rlow appeared at 48.8 ps and its total energy is 2458.37 kcal/mol. The mean square deviation is less than 0.29Å2 during the last 10 ps. Both the loop and TM regions of MC4Rlow model were then minimized to −0.37 × 104 kcal/mol with the rms deviation value ≤0.028 kcal/mol·Å. The calculated distance deviations of C atoms for TMs between the final model of hMC4R and bovine rhodopsin are: TM1: 0.88Å, TM2: 0.72Å, TM3: 0.82Å, TM4: 0.96 Å, TM5: 0.64Å, TM6: 0.89Å, TM7: 0.84Å. The largest deviation is found in TM4.
3.1.3. Features of the Modeling Results This simulation job adopted homology segments, respectively, for TMs and other domains, based on the evidence that the individual segments in GPCRs could fold independently whether for TM or loop domains. It has been reported to simulate other GPCRs by the same strategy (Orry and Wallace, 2000). The 3D structure of the modeled hMC4R molecule is shown in Fig. 3. Profiles_3D program was used to check the correctness of the modeled structure including side-chain environment parameters. The compatibility score of hMC4R is 58.06; it proves that the final structure could be accepted by this homology simulation procedure. The third transmembrane helix is the longest one. The fourth sixth and seventh transmembrane helices are
468699.qxd 7/31/03 10:28 AM Page 341
Structure of the Human Melanocortin 4 Receptor
341
Fig. 5. Predicted docking of the RFF triplet peptide 3D structure to the modeled MC4R molecule. The predicted docking location of RFF triplet (drawn in dark and in stick and ball mode) to the modeled hMC4R molecule (drawn in grey and in a ribbon).
approximately parallel to each other. These features could also be found in the template molecule. Two proline kinks are present at Pro260 (TM6) and Pro299 (TM7) where the two conserved residues also show a kink phenomenon in the template structure. There is an -helix in intracellular loop3 (I3, Arg220-Thr232; Fig. 3). Both secondary structure prediction and structural alignment gave the same structural
frame. The latest experimental datum (Chung et al., 2002) suggested that Arg220 and Thr232 play crucial role in the receptor-G protein interactions, where these two residues were right at the start and end positions of this helical segment. It indicates this -helical segment is a basic motif for G protein coupling with MC4R. A short -helix is in the N-terminal of the cytoplasmic tail. It is conservative among every subtype of
468699.qxd 7/31/03 10:28 AM Page 342
342
Yang, Wang, Dong, Ling, Yang, and Chen
Fig. 6. Detailed view of the hydrophobic cave. The benzene ring B of the ligand (drawn in red and in stick and ball mode) extrudes in a hydrophobic cave of hMC4R (drawn in green and in a ribbon). Benzene ring A is representing the benzenoid form of ligand’s middle phenylalanine, benzene ring B representing another’s.
the MCR family members and clearly distinct from TM7 in the final model (Fig. 3). Secondary structure prediction presented consistent results. A recent experiment declares that this part is crucial for cell surface location (Ho and MacKenzie, 1999); this structure feature indicates that it is a basic motif for melanocortin receptors plasma membrane targeting.
3.2. The Ligand-Receptor Complex Model
located toward the extracellular side of TM domains, mainly including Asp122, Asp126, Phe261, Leu265, etc. And then put the positively charged residue-Arg of the triplet peptide near the more negatively charged region containing two typical negatively charged residues: Asp122 and Asp126 (Fig. 4a). The other two benzene rings are put toward the strong hydrophobic part mainly located near the extrcellular loop2, 3 (Fig. 4b).
3.2.1. Putative Binding Site and Manually Docking
3.2.2. Docking and Refinement of the Ligand-Receptor Complex
For the putative binding site selection, we have first analyzed the hydrophobic surface and electric charge surface of the modeled receptor (Fig. 4). It shows that a series of hydrophobic and negatively charged amino acids
After manually arranging the ligand to its putative binding pocket, Grid/Monte Carlo Minimization docking was run on the initial complex. In this strategy, it provides several random conformations as the seeds during the
468699.qxd 7/31/03 10:28 AM Page 343
Structure of the Human Melanocortin 4 Receptor
343
Fig. 7. Detailed view of the electrostatic interactions and putative H bonds. The arginine of the ligand (drawn in red and in stick and ball mode) show electrostatic interactions with Asp 122 of hMC4R (drawn in green and in a ribbon). Putative H bonds are showing in dash lines.
simulation process. Five potential structures were gained and the lowest energy one was selected for continuous energy minimization till the rms value ≤0.08 kcal/mol·Å. The total energy for the final complex is −0.41 × 104 kcal/mol (some 400 kcal/mol lower than the energy of the receptor without the ligand). And its average deviation of TMs and loop regions from the initial ligand-receptor template is 1.1Å on C atoms. The final ligand geometry did not change much, as the C backbone differred only 0.3Å from the initial ligand structure. Figure 5 shows the location of the ligand in the final ligand-receptor complex.
3.2.3. Interactions of the Ligand with Receptor According to the simulation results, seven transmembrane domains formed the shape of a distended teardrop in vertical view and the triplet peptide was bound to a region flanked by TM3, TM4, TM5, and
TM6 domains (Fig. 5). This location is prevalent in ligand binding sites of GPCRs (Flower, 1999). Analysis of this binding region derived from docking results indicates it mainly contains the following residues: Asp122, Asp126, Ile129, Leu134, Leu133, Leu197, Phe201, Phe261, Phe262, and Phe284. The benzene ring B extrudes in a hydrophobic cave, consisting of Leu133, Leu134, Leu197, Phe201, Phe261, and Phe262 (Fig. 6). We also observed that the arginine of ligand forms electrostatic interactions with the negatively charged Asp122 of TM3. There are still two possible hydrogen bonds between Asp122 and the arginine of ligand (Fig. 7). Furthermore, two -stacks have been observed to be between two pairs of ligand benzene rings and Phe261, due to the respective distances among their benzene rings center are ∼5.7Å and 6.2Å (Fig. 6). These interactions should be another contribution for RFF ligand binding. And hydrogen bond is also possibly formed between the ligand and the side chain of Asp126
468699.qxd 7/31/03 10:28 AM Page 344
344 (Fig. 7). All hydrogen bonds here help to stabilize the ligand binding. In this complex model, we found that extracellular loop1 exhibited no contribution to the ligand binding effects, and TM1 and TM7 domains did not attribute either, but other two extracellular loops participated in the docking process (Fig. 5). This result is consistent with the chimeric receptor data (Yang et al., 1999). There are also many binding pocket residue pairs showing strong interactions between the benzenoid forms: Phe201/Phe262, Phe262/Trp258, Phe261/Phe284, Tyr276/Phe284. It is another characteristic of the receptor that these -stack interactions will reinforce the binding pocket structure. Comparing our model with Prusis’ model, which deduced the agonist binding site of MC1R to be among TM1, TM2, and TM7 domains, our model gives an antagonist binding depiction at a different receptor subtype and can explain the specificity that is observed for inhibitor that shows high binding effects of MC4R but not of MC1R (Yang et al., 1999). Not all of the residues participating in the binding effects will directly interact with the ligand or locate in the putative sites; some of them might mediate interactions with the inhibitor via water molecules. Others might alter the receptor’s whole conformation, but they are not specifically restricted to its binding site when extracelluar ligand binding to the receptor. Situations here are complicated. Our current model could explain a great part of the recent acquired mutations and chimeric receptor datum for RFF ligand binding (Haskell-Luevano et al., 2001; Yang et al., 1999, 2000). In conclusion, we found that the G-protein coupling domain involving residues Arg220 and Thr232 is an helical segment. Our putative identification of the inhibitor binding site, which mainly functions as a hyndrophobic pocket locating among TM3, TM4, TM5, and TM6 domains, can help to design peptides that enhance binding of hMC4R. Furthermore, TM1 and TM7 domains are not involved in inhibitor binding in our research. Thus, mutant forms of the hMC4R with nonconserved substitutions in these domains should still bind to other AGRP analogs which all owned the conserved RFF motif.
ACKNOWLEDGMENTS The work here was supported by National Scientific Foundation of China (Grant Number: 39890070), 863 High Tech Foundation (Grant Number: 2001AA23402106), and
Yang, Wang, Dong, Ling, Yang, and Chen Foundation of Chinese Academy of Sciences (Grant Number: KSCX2-2-07). REFERENCES Baldwin, J. M., Schertler, G. F., and Unger, V. M. (1997). J. Mol. Biol. 272: 144–164. Barsh, G. S., Farooqi, I. S., and O’Rahilly, S. (2000). Nature 404: 644–651. Chen, R. S., Leung, H. W., Dong, Y. C., and Wong, R. N. (1996). J. Protein Chem. 15: 649–657. Chhajlani, V., and Wikberg, J. E. (1992). FEBS Lett. 309: 417–420. Chhajlani, V., Muceniece, R., and Wikberg, J. E. (1993). Biochem. Biophys. Res. Commun. 195: 866–873. Farooqi, I. S., Yeo, G. S., Keogh, J. M., Aminian, S., Jebb, S. A., Butler, G., et al. (2000). J. Clin. Invest. 106: 271–279. Flower, D. R. (1999). Biochim. Biophys. Acta 1422: 207–234. Gantz, I., Miwa, H., Konda, Y., Shimoto, Y., Tashiro, T., Watson, S. J., et al. (1993). J. Biol. Chem. 268: 15147–15149. Gantz, I., Shimoto, Y., Konda, Y., Miwa, H., Dickinson, C. J., and Yamada, T. (1994). Biochem. Biophys. Res. Commun. 200: 1214–1220. Gether, U., and Kobilka, B. K. (1998). J. Biol. Chem. 273: 17979–17982. Haskell-Luevano, C., Cone, R. D., Monck, E. K., and Wan, Y. P. (2001). Biochemistry 40: 6164–6179. Ho, G., and MacKenzie, R. G. (1999). J. Biol. Chem. 274: 35816–35822. Ji, T. H., Grossmann, M., and Ji, I. (1998). J. Biol. Chem. 273: 17299–17302. Kim, C. S., Lee, S. H., Kim, R. Y., Kim, B. J., Li, S. Z., Lee, I. H., et al. (2002). J. Biol. Chem. 277: 31310–31317. Lefkowitz, R. J. (1998). J. Biol. Chem. 273: 18677–18680. Ling, L. J., Shi, X. F., Ni, X. S., and Chen, R. S. (1989). Acta Biophys. Sinica 5: 293–297. McNulty, J. C., Thompson, D. A., Bolin, K. A., Wilken, J., Barsh, G. S., and Millhauser, G. L. (2001). Biochemistry 40: 15520–15527. Mountjoy, K. G., Robbins, L. S., Mortrud, M. T., and Cone, R. D. (1992). Science 257: 1248–1251. Ollmann, M. M., Wilson, B. D., Yang, Y. K., Kerns, J. A., Chen, Y., Gantz, I., et al. (1997). Science 278: 135–138. Orry, A. J., and Wallace, B. A. (2000). Biophys. J. 79: 3083–3094. Palczewski, K., Kumasaka, T., Hori, T., Behnke, C. A., Motoshima, H., Fox, B. A., et al. (2000). Science 289: 739–745. Prusis, P., Frandberg, P. A., Muceniece, R., Kalvinsh, I., and Wikberg, J. E. (1995). Biochem. Biophys. Res. Commun. 210: 205–210. Prusis, P., Schioth, H. B., Muceniece, R., Herzyk, P., Afshar, M., Hubbard, R. E., et al. (1997). J. Mol. Graph. Model. 15: 307–317. Roselli-Rehfuss, L., Mountjoy, K. G., Robbins, L. S., Mortrud, M. T., Low, M. J., Tatro, J. B., et al. (1993). Proc. Natl. Acad. Sci. USA 90: 8856–8860. Schioth, H. B., Muceniece, R., and Wikberg, J. E. (1996). Pharmacol. Toxicol. 79: 161–165. Schwartz, M. W., Woods, S. C., Porte, D., Jr., Seeley, R. J., and Baskin, D. G. (2000). Nature 404: 661–671. Tota, M. R., Smith, T. S., Mao, C., MacNeil, T., Mosley, R. T., Van der Ploeg, L. H., et al. (1999). Biochemistry 38: 897–904. Vaisse, C., Clement, K., Guy-Grand, B., and Froguel, P. (1998). Nat. Genet. 20: 113–114. Yang, Y. K., Dickinson, C. J., Zeng, Q., Li, J. Y., and Thompson, D. A. (1999). J. Biol. Chem. 274: 14100–14106. Yang, Y. K., Fong, T. M., Dickinson, C. J., Mao, C., Li, J. Y., Tota, M. R., et al. (2000). Biochemistry 39: 14900–14911. Yeo, G. S., Farooqi, I. S., Aminian, S., Halsall, D. J., Stanhope, R. G., and O’Rahilly, S. (1998). Nat. Genet. 20: 111–112. Zhu, L. L., Hou, T. J., Chen, L. R., and Xu, Y. J. (2002). Acta Chimi. Sin. 60: 43–48.
